Majority and Minority Ethnic Voting in New Democracies

Home , Albanian Americans, Bosnian Party, Elections in Latvia, Ethnic enclave, Ethnocracy, Frank Aarebrot

Identity and Agency: Majority and Minority Ethnic Voting in New Democracies

Benjamin P. McClelland

Submitted in partial fulﬁllment of the requirements for the degree of Doctor of Philosophy under the Executive Committee of the Graduate School of Arts and Sciences

COLUMBIA UNIVERSITY

Benjamin P. McClelland

This dissertation examines how ethnic identities are politicized through elections in new democracies. Using the cases of post-communist Latvia and Bosnia and Herzegovina, I compare the electoral success of campaigns which appeal to voters on the basis of ethnicity to those do not. I argue that ethnic parties are most likely in groups for whom two conditions are met. First, ethnicity must meaningfully differentiate ethnic insiders from outsiders, in such a way that voters will believe policy benefits will likely result from political representation for the group. Second, electoral institutions must ensure that the political mobilization of the group will result in electoral victory. These two conditions create fundamentally different incentives for ethnic majority groups and ethnic minority groups simply because of differences in group size. In most democracies with a large minority population, ethnic voting will be more likely among the majority group than the minority group, unless institutions encourage minority group voting by lowering barriers to entry. The results demonstrate the qualitatively different ways groups use ethnic identities as a resource to achieve political objectives, with important implications for minority group representation, political participation, and democratic governance in diverse societies. Contents

1 Introduction 1 1.1 Why Study Ethnic Voting? ...... 4 1.2 A Theory of Ethnicity as a Political Asset ...... 6 1.2.1 Ethnicity as Informational ...... 7 1.2.2 Ethnic Identity as a Political Resource ...... 9 1.2.3 Majority and Minority Group Dynamics ...... 12 1.3 Contribution to Existing Research ...... 15 1.4 Plan of this Dissertation ...... 20

2 A Theory of Group Size and Ethnic Voting 22 2.1 Deﬁnitions ...... 23 2.1.1 Identity ...... 23 2.1.2 Ethnicity ...... 26 2.2 Linking Ethnicity and Voting ...... 29 2.2.1 Ethnicity and the Challenges of Elections ...... 29 2.2.2 The Beneﬁts of Ethnic Representation ...... 34 2.2.3 Ethnicity as Coordination Device ...... 39 2.3 Institutions and Group Status ...... 42 2.4 Visualizing the Theory ...... 45 2.4.1 Ethnic Voting Decisions in a Very Small Polity ...... 46 2.4.2 Ethnic Voting in Reality: Albanians and the Albanian Diaspora . . . . 50 2.5 Conclusion ...... 55

3 Research Design 57 3.1 Challenges and Approaches to Empirical Veriﬁcation ...... 58 3.2 Features of this Design ...... 66 3.2.1 Data Sources ...... 70 3.2.2 Mixed Methods ...... 76

4 Ethnic Politics in Bosnia and Herzegovina 78 4.1 Historical Background ...... 79 4.2 The Wartime Origins of Bosnian Democratic Institutions ...... 86 4.3 The Evolution of Bosnian Political Actors under the New System ...... 90 4.4 The Political Use of Ethnicity in Bosnian Elections ...... 93 4.4.1 Party Names ...... 94 4.4.2 Party Logos ...... 98 4.4.3 Campaign Rhetoric ...... 102 4.5 Classifying Bosnian Political Parties ...... 113

i 4.5.1 A Three-Part Typology ...... 113 4.5.2 Can a Bosnian candidate really be “non-ethnic?” ...... 115 4.6 Conclusion ...... 119

5 Ethnic Voting in Bosnia and Herzegovina 121 5.1 Split-Ticket Voting and Ethnicity ...... 122 5.2 Data Collection and Sources ...... 127 5.2.1 Coding Parties ...... 128 5.2.2 Voting Data ...... 130 5.3 Analysis ...... 132 5.3.1 Descriptive statistics ...... 132 5.3.2 Local majority/minority status as independent variable ...... 135 5.3.3 State level demographic status as independent variable ...... 139 5.4 The Results in a Single City: the Case of Odžak ...... 143 5.5 Variation over Time ...... 145 5.6 Conclusion ...... 150 5.7 Appendix: Robustness to Alternate Coding: Voter Perceptions over Candidate Declarations ...... 151

6 Ethnic Politics in Latvia 154 6.1 Historical Background ...... 155 6.2 From Soviet Republic to European Democracy ...... 161 6.3 The Political Use of Ethnicity in Latvian Elections ...... 171 6.3.1 Party Names ...... 173 6.3.2 Campaign rhetoric ...... 178 6.3.3 A Note on Party Logos ...... 190 6.4 Classifying Latvian Political Parties ...... 194 6.4.1 A Three-Part Typology ...... 194 6.4.2 Can a Latvian Party Really be “non-ethnic?” ...... 195 6.5 Conclusion ...... 198

7 Latvian Elections over Time 200 7.1 Over-Time Variation in Latvia ...... 201 7.2 The Data Set ...... 206 7.3 The General Trajectory of Latvian Party Consolidation ...... 210 7.3.1 Party Vote Share over Time ...... 210 7.3.2 Candidate-Level Analysis ...... 213 7.4 Ethnic Position Change in Practice ...... 222 7.5 Conclusion ...... 228

8 Conclusion 230 8.1 Implications for Future Studies ...... 233 8.2 The Fate of Diverse Democracies ...... 240

References 249

ii List of Figures

2.1 Viable Yellow Party; Non-Viable Blue Party ...... 47 2.2 Two Electoral Districts Increases Viability for Blue Party ...... 47 2.3 Greater Diversity Increases Beneﬁts from Yellow Party ...... 47

4.1 Bosnian Eﬀective Number of Parties 1996-2014 ...... 91 4.2 Bosnian Ethnic Party Logos ...... 99 4.3 Bosnian Non-Ethnic Party Logos ...... 100 4.4 Bosnian Campaign Advertisements ...... 106

5.1 Ethnic Voting in Bosnia at Cantonal and State Levels ...... 133 5.2 Ethnic Voting in Bosnia by Group Type ...... 134 5.3 Ethnic Voting Scatterplot ...... 135 5.4 Comparison by Group ...... 140 5.5 Ethnic Vote Share Precinct Distribution ...... 141 5.6 Bosnian SNSD Logo, adopted 1996 ...... 145 5.7 Bosnian SNSD Homepage, February 2017 ...... 147 5.8 Campaign Posters for Bosnian SBB ...... 148 5.9 Logos for Bosnian BPS ...... 149

6.1 Naturalizations in Latvia: 1995-2018 ...... 171 6.2 Latvian Mainstream Party Logos ...... 190 6.3 Latvian State Symbols: Established 1923 & 1921 ...... 192 6.4 Ethnic Latvian & Far Right Parties ...... 193

7.1 Vote Share by Electoral Appeals in Latvia: 1995-2014 ...... 211 7.2 Probability of Election in Latvia by Appeal Type over Time ...... 216 7.3 Likelihood of Candidate Changing Ethnic Appeal in Latvia 1998-2014 . . . . . 218 7.4 Count of Candidates who Changed Ethnic Identiﬁcation between Elections in Latvia 1998-2014 ...... 220 7.5 New Candidate Entry in Latvia by Appeal Type 1998-2014 ...... 221

iii List of Tables

2.1 Two Components of Ethnic Voting ...... 45

4.1 Bosnian Party Typology ...... 115

5.1 Institutional and Demographic Diﬀerences between Cantonal Assemblies and House of Representatives ...... 124 5.2 Hypotheses ...... 127 5.3 Split-Ticket Voting Regression Results ...... 137 5.4 Disaggregated Dependent Variable ...... 139 5.5 Voting Returns from Odžak 2010* ...... 143 5.6 Robustness to Alternative Codings ...... 152

6.1 Latvian Party Typology ...... 195

7.1 Hypotheses ...... 206

iv 1

Introduction

On my first ever visit to the former Yugoslavia, I found myself at a bar in Sarajevo talking to a fellow patron. He was roughly my age, but we had clearly lived very different lives. He was still in primary school when the Bosnian War had begun, and he had spent a substantial chunk of his childhood in a war zone. The experience had filled him with contempt and disgust that had not abated in the years following the war. He said that the world thought there were three types of people in Bosnia, but that was wrong. Yes, there were Bosniaks, Croats, and Serbs, but there was also a fourth type: “normal people.” His dream, he said, was to send all the Bosniaks to Turkey, and the Croats to Croatia, and all the Serbs to Serbia, and then Bosnia could be a great country, full of “normal” people. Those people, he thought, would probably have a pretty good time with everyone else gone. My thought at the time was that this person must be extremely atypical. People who hang out in bars in the capital to practice their English with foreigners are hardly a representative sample of Bosnians. Ethnic divisions are so deeply entrenched in virtually all aspects of Bosnian society that to imply that the overwhelming majority of Bosnians who strongly identify with one of the three main ethnic groups are somehow not “normal” can hardly be a popular position. Ethnic segregation is also the norm in the political life of the country, and parties representing those three main ethnic groups had been the dominant political players since the end of the civil war. Out of curiosity, I asked my companion who he planned to vote for in the next election. “The HDZ,” he answered, giving the Bosnian abbreviation for the Croatian Democratic Union, one

1 of those very same ethnic parties. I told him his answer surprised me. Didn’t he know that the HDZ justified its existence by representing one of those ethnic groups of “not normal” people? Had he forgotten that the HDZ leadership led the combatants in the civil war that had destroyed his city and his childhood? Had he heard the party’s candidates on TV, declaring that their goals were greater autonomy and more rights exclusively for Croats? Of course he knew all this, much better than I did. But he and his family were Croats, he said, and Croats were the smallest group in the country. He knew that the other two groups were electing their own representatives, and if the Croats didn’t take care of themselves, no one else would. Croats needed jobs and social services, too, and a Croat party was the only way to provide that. He didn’t see his choice as hypocritical or disingenuous. It was simply what he had to do, living in a country filled with so many “not normal” people. This person saw his primary identity in non-ethnic terms, and saw the ethnic division of society as undesirable. His ethnicity was something he had thought about a lot, but did not regard positively. He associated ethnicity with violence and childhood trauma, and saw people who thought of themselves primarily as members of an ethnic community as almost pathological. Yet when voting he helped to maintain and reinforce those ethnic divisions. He supported a party that campaigned to help a specific ethnic group even at the expense of others with whom he felt an affinity. His was a decision to support an ethnic political party even while he found the very idea of ethnic politics objectionable. That this person should vote for an ethnic party is surprising to say the least. But in a way, it’s surprising that anyone votes for an ethnic party. Ethnic identities are the result of inheritance, not choice. They are arbitrarily assigned at birth, a quirk of history and happenstance, usually beyond the ability of any individual to change. It’s surprising that people would attach so much personal significance to something completely beyond their control. Ethnicity seems almost capricious when compared to other social divisions and identity categories like age, class, or gender. While a rich voter and a poor voter are readily identifiable by differences in what they possess, a Croat voter and a Serb voter, for example, are often only accepted as different because of historical consensus that they are different. Yet ethnicity is hugely consequential for voting in much of the world. In both developing and developed countries, ethnic divides strongly

2 correspond to vote choice. My Balkan barroom colleague suggests one under-appreciated reason as to why this may be the case. His decision to support an ethnic party was not the result of the way he felt about himself or his group, but instead the context in which he found himself and his perceptions of the likely behavior of others. The practical reasons to support an ethnic party were so compelling that they overcame the disdain he felt for the ideology of the people was voting for. He didn’t like it, but he still chose to support an ethnic party. This dissertation asks why voters in divided societies vote for ethnic parties. To answer that question, I examine ethnic voting as a conscious and calculated decision to use identity as a resource to gain tangible benefits from the state. I propose that ethnic voting is most likely when ethnic mobilization represents a viable path to controlling policy outcomes. To understand why people choose to vote for someone who claims to represent their ethnic group, we should first understand what they can expect to get by putting such a person in a position of power, and why they think that person can deliver on those promises. Ethnic identities can be a political resource, useful both for voters concerned with creating favorable state policies, and for would-be politicians concerned about winning office. But not all ethnic identities are equally useful for pursuing these goals. For some ethnic groups, campaigns built on ethnic identities reliably help voters enact favorable policies and candidates win elections. For others, ethnic appeals are routinely defeated by ideological campaigns or other types of parties which seem more appealing. In order to determine why some voters support ethnically-defined political parties, we have to understand what exactly it is that ethnic voting offers. To do so, I look at two specific components: the potential benefits to the group of ethnic representation, and the likelihood that such group mobilization will result in electoral victory. These conditions do not exist for all groups equally. In fact, they do not even exist for all groups within the same country. The incentives to support ethnic parties can vary enormously between groups even within the same country operating under the same rules. Specifically, the distinction between majority and minority groups is enormously important. Contrary to many expectations, this dissertation argues that ethnic majority groups are most likely to win elections by campaigning on ethnic appeals, not minority groups. Democratic elections rely on the principal of majority rule, and in most cases ethnic majority groups are more likely to win by

3 politicizing ethnicity than minority groups. Ethnic minority groups can and do successfully campaign on ethnic issues, but this is most likely to happen in institutional contexts where minority groups enjoy the benefits of being majorities through segregation into their own discrete units. Ethnic voting is not governed by the same rules among minority and majority group members, because the strategic calculations facing the two groups are fundamentally different. Ma- jority groups may have the greatest likelihood of winning political office through ethnic voting, but usually have the least to gain through ethnic representation. For minority groups, the situation is reversed: they may have the most to gain through being represented in government, but will also have the hardest time winning office because of their reduced numbers. To explain ethnic voting, it is important not only to acknowledge this trade-off between the motive and the means to elect ethnic representatives, but also to understand how electoral rules and political institutions can change the balance between these two often contradictory forces. In the chapters that follow, I seek to explain the phenomenon of ethnic voting, drawing special attention to countries divided between majority and minority ethnic groups.

1.1 Why Study Ethnic Voting?

In much of the world ethnicity is an important predictor of voting choice. Horowitz’s (1985) “ethnic census” voting, under which voting behavior perfectly corresponds to ethnic identities such that election returns resemble census data, is the norm in much of the world. Empirical studies have documented widespread ethnic voting patterns in much of the world including Africa,1 Eastern Europe,2 South Asia,3and the long-standing democracies of Western Europe and North America.4 This pattern raises normative concerns about ethnicity subverting democratic accountability. The tendency to support coethnics may be so strong as to overwhelm voters’ capacity to assess candidates’ competence or policy positions. The normative value of democracy rests on the way

1See Nugent (2001), Bratton & Kimenyi (2008), Eifert, Miguel & Posner (2010), Ferree (2006). 2See Birnir (2007). 3See Aspinall (2011). 4See Barreto, Segura & Woods (2004), Hill, Moreno & Cue (2001)

4 in which it facilitates positive outcomes through voters’ willingness to “throw the bums out.” If politicians can rely on voters’ support by virtue of their ethnicity, they may be more likely to engage in—and get away with—all manner of corruption, venality, and idiocy. This has led some to conclude that democracy may simply be more difficult in ethnically diverse societies. More diverse countries tend to be backwards economically,5 more prone to conflict,6 and more poorly governed7 than more homogeneous countries. For some, this intu- ition is confirmed by a cursory glance at the quality of life around the world: Nordic European countries like Denmark and Norway which routinely rank highest on quality of life indicators are noticeably ethnically and culturally homogeneous, while those countries plagued by poverty and perennial warfare like Uganda and Afghanistan are among the most ethnically diverse in the world. Increasing diversity in the developed world due to migration has spurned a populist and at times anti-democratic backlash, further supporting this suspicion in the eyes of many observers. The inevitable conclusion is pessimism about the prospects for democracy in the world of the twenty-first century. As much of the developing world is highly ethnically diverse, and the developing world is increasing in its heterogeneity, democracy may very well be headed towards a period of global stagnation or decline. But before accepting such a bleak conclusion, we should look more closely at the mechanisms which link ethnic diversity to political outcomes. The causal chain through which diversity affects democracy is held together by the decisions of individual societal actors. Identities are not actors or actions in and of themselves, but rather characteristics that influence outcomes by the ways they structure the context and the background conditions under which actors make decisions. Political and economic outcomes are the accumulated results of decisions made by atomized citizens each acting in what they perceived to be their own best interests. If diversity does make it harder for democracy to survive, it will operate—and thus should be observable– through the decisions of political actors. In this dissertation I concentrate on ethnic voting. The ability of citizens to choose their leaders is the most important characteristic of democracy, and the election of leaders is the first

5See Easterly & Levine (1997) and Alesina, Baqir & Easterly (1999). 6See Fearon & Laitin (2003), and Montalvo & Reynal-Querol (2005). 7See La Porta, Lopez-de Silanes, Shleifer & Vishny (1999).

5 step in any democratic policy-making. To understand how democracy can function in ethnically diverse societies, I explore how voting decisions are made under conditions of ethnic diversity. Election results are the product of interactive decision making made by two groups of actors: voters and politicians. When ethnic identities are politicized, it is because both sets of actors have chosen to politicize them. In other words, when ethnicity is politically relevant, it is because elites have chosen to mobilize voters on the basis of ethnic identities, and voters have chosen to support candidates who do so. What is necessary then is a theory which links ethnicity and voting through the mechanism of human agency. Why do politicians choose to campaign on ethnic platforms? What strategies are available to politicians who choose to politicize ethnicity, and why do voters support candidates who choose to do so? In asking this question, I hope not only to further our collective understanding of elections, democracy, and political development, but also to explore the role of ethnicity and identity in the human experience. Questioning the link between individual-level ethnic identities and group-level political mobilization ultimately speaks to fundamental issues regarding the role of the individual in society, and how an individual’s experience is integrated into a societal whole. This dissertation thus seeks not only to explain political outcomes and understand the prospects of democratic governance in diverse societies, but contribute to the understanding of the human condition and the relationship between individuals and the societal and political communities in which we ﬁnd ourselves.

1.2 A Theory of Ethnicity as a Political Asset

This dissertation argues that ethnicity should be understood as a political resource possessed both by voters who want to receive tangible benefits from the state, and by politicians who want to win office. Ethnic voting is most likely when the value of this resources is high for both sets of people: voters want ethnic representation in the state because individuals stand to gain when state resources are redistributed along ethnic lines and politicians can use ethnic identities as a way to win political office. This is not the case in all circumstances, however. To understand the value of ethnic representation to these two sets of people, it is important to understand how

6 ethnic identities inform voting decisions.

1.2.1 Ethnicity as Informational

An ethnic identity is an individual identity based on personal attributes associated with decent. Ethnic identities are both highly visible, and difficult to change.8 These characteristics make ethnic identities useful as a heuristic device. During an electoral campaign, a politician must reach out to potential voters, and persuade those voters that they should vote for her over her rivals. Since campaigning is ultimately about persuasion, mobilization, and coordination, the way in which the visibility of ethnic identities present information about candidates to voters may represent a politically useful way for building linkages between would-be leaders and their possible constituents. One potential stream of information conveyed by ethnicity is the identification of commonality. By its very nature, identity establishes the personal characteristics that make one either distinct from or similar to others around her. It is the intellectual, philosophical, and cognitive tool that humans use to distinguish self versus other, sameness versus change, and unity versus diversity (Lagasse 2017). All identity categories ranging from the profound (religion, sexuality, race) to the frivolous (Arsenal supporter, birdwatcher, Star Wars fan) help to fulfill this function, but ethnic identities are unique. Since ethnic markers are highly visible and difficult to change, they are extremely useful for the immediate and efficient identification of in-groups and out-groups. For many ethnic groups, ethnicity is communicated almost immediately by appearance, language, accent, dress, or mannerism. Voters making voting decisions in a limited information environment can readily gauge which candidates are members of their own ethnic group, sorting options into ethnic insiders and outsiders. Just as ethnicity can be useful to voters for deriving information on candidates, it can also be useful to candidates in deriving information about voters. Public opinions on specific political issues may change fairly drastically in the short term, and be difficult to assess with certainty. Ideology and partisanship are rarely visually observable, and can require costly information gathering for politicians seeking to appeal to constituents on those grounds. Ethnic identities,

8This deﬁnition borrows heavily from Chandra (2004). For a full discussion of the deﬁnitions of identities used in this dissertation, see Chapter 2.

7 on the other hand, remain relatively constant in the short- and medium-term, and are often easily observable with minimal effort. How dominant a specific group is within a specific area is usually easily assessed and slow to change, and candidates can gauge potential ethnic backgrounds of voting communities fairly easily. In this way, ethnic identity is one of the easier forms of information for would be-leaders to obtain about their potential supporters. The fact that ethnic identities can be useful for both politicians and voters makes them important in building linkages between the two. Voters care about electing leaders who will enact policies that will benefit them. They seek to identify politicians who share their interests and priorities, and cast their votes in a way that will help elect those leaders. By relying on ethnic cues, voters can at the very least identify which candidates share their descent-based characteristics, and which do not. Politicians, on the other hand, care about getting elected. Democracy is a system of majority rule—the biggest political faction is usually the one that wins. When deciding which electoral group to target, a politician ultimately wants to win over a constituency that is big enough to win office. Ethnic identities help voters gauge the relative size of ethnic communities, and identify potential winners and likely losers before campaigning. But simply because ethnicity conveys information doesn’t mean it always facilitates ethnic voting. Since ethnicity does not directly measure competence, policy position, or intentions, voters may observe ethnicity but ultimately deem other indicators of candidate quality more important. Likewise, politicians may use ethnic identities to determine that even full mobilization of an ethnic community is not likely to produce a big enough base of support to win office. In either of these cases, voters and politicians may be using the information conveyed by ethnic cues to decide not to vote along ethnic lines. Ethnic identities serve an information role, and are therefore not deterministic over individual actions. Whether the information conveyed by ethnicity encourages or discourages voting along ethnic lines depends on the circumstances of the group and the context in which they find themselves. Since this will vary by country, and even by group within the same country, some ethnic identities may be better at mobilizing voters than others.

8 1.2.2 Ethnic Identity as a Political Resource

The word “election” comes from the Latin word for “choice.” While this usually refers to the choice of voters over who will represent them in office, politicians also make choices during electoral campaigns. Politicians get to choose how they present themselves to voters, and what aspects of their own personal brand they will emphasize. Not all labels are equally valuable: most candidates want to be associated in the minds of voters with terms like “honest,” “competent,” “fair,” etc. Most candidates also want to avoid toxic labels that voters find unappealing: “philanderer,” “tax cheat,” or “problem drinker.” Many labels politicians campaign on appeal only to a subset of the electorate, and are not valued equally by society as a whole. In the classic interpretation of democracy, electoral competition stabilizes around a set of symmetrical identity categories in the sense that one party campaigns on one platform, while another campaigns on its opposite. These are the archetypal social cleavages around which elections have historically been structured in the western world: rural versus urban, religious versus secular, and capital versus labor (Lipset & Rokkan 1967). Neither a voter nor a politician can be in both categories at once. If voters are presented with a politician claiming to represent the interests of the rural class and a politician claiming to represent the urban classes, presumably the choice for voters is clear: rural voters support the rural party and urban voters support the urban party. Ethnic categories are different. Ethnic markers, as opposed to ideological or valence categories, cannot be changed, and often cannot be easily hidden. A politician can claim to be for the working class when she is in fact motivated by corporate greed (i.e., an appeal to voters of a “laborer” identity when the true identity is “capitalist”), and such a claim may still be plausible to many voters. Such a façade is harder to present with regard to ethnic identities, as such identities are usually readily apparent through dress, name and custom. But the trade-off for this reliability is a lack of specificity. Ethnic identities in and of themselves do not necessarily provide information on the way in which a politician will direct the policies of the state. A politician claiming the identity of “liberal,” for instance, is basing the individual identity on a set of values that have direct implications on how she feels the state should govern. When this politician makes such an appeal to voters, there are two possible outcomes: either this politician

9 is lying, or she will endeavor to enact policies that support the free market, property rights, and limited government intervention once elected. For certain groups, ethnic representation is a reliable way to obtain real policy benefits from the state. Having a candidate who claims to support coethnics—even independent of ideology, competence, or other candidate quality—is still a highly desirable trait. In this sense, the identity itself is a valuable resource. It helps voters get what they want from the state, and it helps politicians mobilize those voters to win office. The identity is a resource that helps actors pursue their interests. To understand when voters support ethnic appeals over other appeals, it is important to understand which ethnic identities make for the most appealing bases of electoral mobilization. When the expected returns of ethnic representation are high, the value of that identity is high, just as when the utility of ethnic voting is low, the value of that identity is low. If the usefulness of a specific ethnic identity is of a very low value, then a politician campaigning on that basis is likely to be defeated by a rival competing on some other political identity category, be it ideological, regional, or something else. If that value is high, then ethnic appeals are most likely to be supported by voters even when challenged by other identity appeals. There are two factors which affect the likelihood of ethnic voting. The first is the nature of the benefits that would likely flow to the group members from ethnic representation. The second is the likelihood of victory of that group in an election. The reasons that individuals may prefer to have themselves represented on the basis of their ethnic identity fall into three categories. First of all, coethnic representation has certain intrinsic benefits. Individuals who have a strong group attachment may feel a sense of validation and increased self-esteem from seeing members of their own group in power. The prestige that comes with elected office may be viewed positively as an indication of the quality or respect due to the entire group, and voters may think that electing a coethnic to office is worthwhile simply because of the intrinsic psychological benefits that come with victory. Secondly, since ethnic identities correspond to descent-based attributes, they may be correlated with other descent- based characteristics that have immediate bearing on policy preferences. Since these characteristics include religion, language, or cultural practices, ethnicity may strongly predict common

10 interest in policies that protect the ability of ethnic communities to pass down their way of life. Finally, ethnicity could serve as the basis of between-group redistribution and policies of ethnic favoritism. In situations where the state may be in a position to discriminate in the provision of state services or redistribute wealth, having a coethnic in power may be valuable even independent of their stated policy beliefs or competence as a public official, simply because ethnic favoritism will result in state largesse finding its way to the voters. In any of these situations, voting for someone on the basis of ethnicity is a rational decision. It is not necessarily the case that a strong sense of attachment to an ethnic identity is encouraging a thoughtless reflexive decision to support a candidate on the basis of ethnicity, but instead that voters are consciously choosing to support in light of real and tangible pay-offs they receive by doing so. Even if there are reasons why voters may desire political representation for their ethnic group, there is no guarantee that they can have it. For an ethnic identity to be valuable as a means of attaining office, the group has to be able to win elections. The most important determinants here are the size of the group and the electoral context. Take, as an example, a very small group representing one-tenth of the population. In a majoritarian winner-take-all election, this group is hardly likely to win. Even assuming that every member of the group decides to vote for the same candidate–an unlikely scenario given the challenges of group coordination and heterogeneous preferences–one-tenth of the vote share is usually not enough to win political power through democratic contestation. Group size is thus directly related to the usefulness of group identities in obtaining political power, because demographics place a ceiling on the influence a group can have.9 A group representing sixty percent of the total population would not have this problem, since they are already a clear majority. While this same constraint exists for all political actors, ethnic identities differ from other bases of mobilization because of visibility and permanence. Political entrepreneurs generally know in advance what the electoral rules are, and how big a potential ethnic constituency is. As such, they can generally know in advance which groups are

9That’s not to say that the group is powerless. There are several scenarios in which a very small group can wield oversized inﬂuence. Other, larger groups could face more serious coordination dilemmas and split the vote amongst themselves, allowing a more organized and uniﬁed smaller group to win. Or small groups can act as important “king makers” in coalition politics, determining who actually is able to form a government. These situations are far from common, though, and when they do occur it is more often due to the actions of those taken outside the group rather than in it. The fact remains that democratic elections make it harder for smaller groups to become politically powerful. If these situations obtain, it is due to factors outside the control of those in the group itself.

11 likely to be able to win, and which groups aren’t. If a group is large enough, then the value of that ethnicity is much higher than if it is small. This theoretical framework allows us to think about ethnic voting in a way that returns a degree of agency to voters and political elites in ethnically divided societies. Decades of constructivist scholarship have convincingly demonstrated that ethnic identities can be socially con- structed and politically malleable, evolving and either gaining or losing political significance over time. Yet these are generally thought to be the result of long-term structural changes: the product of changes in economic production, colonial policies, or intellectual and cultural movements spanning decades. By looking specifically at the short-term gains when faced with a discrete decision–who to vote for– we can understand the way that individual identity repertoires influence voters’ short-term decision making and structure political outcomes. Voters are unlikely to support identity-driven platforms that offer them little in return, keeping with the constructivist insight that ethnic identities do not take on political significance simply by virtue of their existence. By the same token, political elites are unlikely to base their campaigns on platforms that will deliver few benefits to their constituents and make it easier for competitors to offer more. This means that not all ethnic identities are equally useful as political tools, and the degree to which we should expect voters to respond to ethnic appeals is a function of these two explanatory concepts. There are many reasons groups could vary in the degree of benefits they could derive from state policy, and their likelihood to win office in electoral contestation. One of the most important ways is simple demographics–specifically, the group’s status as either the majority or the minority group.

1.2.3 Majority and Minority Group Dynamics

The discussion above suggests that ethnic voting should not be equally common in societies divided between a demographically dominant ethnic majority group, and a numerically smaller ethnic minority group. This is mainly because while both the benefits of ethnic representation and the likelihood of winning office are partially determined by group size, the two factors push in different directions for different groups. Ethnic majority groups are likely to have the largest

12 chance of winning office through ethnic voting, but the least to gain if they win. For ethnic minority groups, the situation is the opposite: they would likely benefit the most from ethnic representation, but have the least chance of winning. Take, for instance, a voter who is a member of the dominant ethnic group in a country with near universal ethnic homogeneity—as is the case for almost all voters in nearly ethnically uniform countries like South Korea, Swaziland, or Tunisia. In these cases, electing a coethnic is basically guaranteed. Voters can be highly confident that their ethnic group is large enough that ethnic mobilization is guaranteed to produce a winning candidate, as virtually all voters have the same ethnic identity. The degree to which ethnic identities broadcast relative group strength is so high, that ethnic voting is clearly a viable path to victory. Yet in this scenario, ethnic identities are likely to have little to no overlap with material interest. Since everyone is the same, ethnic identities serve no useful basis for redistribution. For a candidate to campaign on a platform of ethnic identity is essentially to appeal to everyone, and such an appeal is likely easily defeated by a more narrowly defined policy appeal on the basis of ideology, region, class, etc. Thus for a voter in this situation, the probability of winning under ethnic mobilization is extremely high, but the benefits of explicitly ethnic representation are virtually nil. This is a very different situation from that of a member of a majority ethnic group in a more ethnically diverse society, as is the case with white voters in the United States, or Malays in Malaysia. Voters in these groups know that they constitute a substantial majority of the population, and that ethnic mobilization is therefore likely to be a viable path to victory. But as opposed to the dominant group members in the purely homogeneous society, these voters have more to gain by ethnic representation. Since there are minority ethnic group members, these groups would be able to gain by taxing ethnic outsiders and then redistributing to coethnics. They may be able to exclude ethnic minorities from labor competition through discriminatory laws that advantage the ethnic majority group, or grant their own cultural practices or language privileged positions against those of the minority community. Under these conditions, the value of ethnic representation has gone up. Voters in this category should be more likely to support ethnic appeals from politicians than those in the purely homogeneous context, because they have more to gain by ethnic representation. The choice between an ethnic appeal and an appeal to

13 some other cleavage is now more complicated, because an appeal on class, region, or ideological lines is likely to cross-cut the ethnic divide, and include members of the ethnic out-group. These voters have a high likelihood of winning elections, and also have more to gain from ethnic representation than they would if they lived in completely homogeneous country. Ethnic voting is therefore more likely for these majority-group voters in diverse countries. For minority groups such as Hungarians in Romania, Arabs in Israel, or Maori¯ in New Zealand, the overlap between interest and ethnicity is likely to be highest. As a small group, installing themselves in power and then taxing ethnic outsiders to redistribute benefits to themselves would result in a massive windfall for group members. Beyond material gains, these groups are also likely vulnerable to expropriation, discrimination, or irrelevance at the hands of numerical majorities. Informally, they may find it difficult to perpetuate cultural practices or speak minority languages in an environment where absorption and assimilation into larger groups is practically expedient. Their languages and ways of life may be legally threatened by the policies of a democratically elected parliament dominated by a majority group either hostile or indifferent to needs of the minority. Given their propensity to be overwhelmed by simple numerical domination, they have the most to gain from state action. Thus the benefits of ethnic representation are likely to be extremely high. But because they are numerically small, these groups face very high barriers to entry for political power. Even if they are able to mobilize the entire group behind a single candidate in a display of ethnic solidarity and voting unity, their overall vote share will be capped by their share of the population. Their low likelihood of actually winning office offsets the potential benefits to be gained, resulting in a low likelihood of ethnic voting. Majority and minority groups thus face different challenges and incentives: majority groups have a high probability of winning, but a low overall payoff under ethnic voting. Minority groups have a low probably of winning, but much to gain under ethnic voting. As group size increases, so does the likelihood of winning elections. But at the same time, increasing the size of the group decreases the pay-off of ethnic representation. This suggests that both very large majority groups and very small minority groups have relatively low incentives to support ethnic platforms, albeit for very different reasons.

14 Starting from a framework that ethnic voting behavior is best understood using the two dimensions of commonality of interest and likelihood of victory, this dissertation argues that group status as either a majority or a minority group is one of the most important determinants of individuals’ likely support for ethnically-delineated party platforms. This is not to say that there are not other reasons that ethnic groups could vary on these two dimensions. A group’s commonality of interest can change due to between-group inequality, historical animosity, or a group’s idiosyncratic understanding of its own identity and history. Likewise, the likelihood of winning for a group can vary enormously based on institutional circumstances in which the group finds itself. Proportional electoral systems, bicameralism, decentralization, federalism, regional autonomy, and territorial districting are all common institutional features of ethnically divided countries, often implemented with the expressed purpose of facilitating political representation of diverse identity groups. Nevertheless, I argue that all of these factors which influence ethnic voting behavior operate through these two dimensions of interest and viability. Majority/minority status exerts a strong influence on these dimensions which is of extreme importance in understanding the relationship between identity and elections.

1.3 Contribution to Existing Research

This dissertation seeks to explain why and how people make the decisions which turn ethnic cleavages in society into political cleavages in parliament. This issue is not only one of scholarly interest, but also one of great normative and policy importance. Ethnic diversity can be highly destabilizing and problematic for democratic governance. How to mitigate these problems has been one of the most contentious issues in political science in recent decades. Much of the disagreement is the result on the respective assumptions made by the various camps on the reasons people decide to vote their ethnicity. The now decades- old debate between consociationalists10 and centripetalists11 in many ways revolves around this question. Consociatlionists assume that ethnic diversity presents a problem for democracy because the tendency to vote one’s communal interests is so strong as to be essentially insur-

10See Lijphart (1977), McGarry & O’Leary (2006), and Nordlinger (1972). 11See Horowitz (1985) and Reilly (2001).

15 mountable, and thus conclude that institutions should be purposefully designed to facilitate the representation of those communal identities in order to make democratic governance possible. Centripetalists, on the other hand, argue that institutions can be designed to foster inter-ethnic coalitions that downplay the ethnic cleavage in favor of broader civic political bases of mobilization. Encouraging people to vote along non-sectarian lines creates larger political bases of support more conducive to stability and public goods provision, making a multiethnic democracy possible. The main difference between these two camps is in the strength they assume identity has over individual behavior. Consociationalists believe that identities are stronger than institutions –and that if institutions are not designed to empower those identity communities, the most likely result is a revolt against those same institutions. Centripetalists believe that if institutions disadvantage identity representation, voters will find some other basis on which to mobilize. I suggest that the way to resolve this debate is to refocus on individual incentives. Identities and institution are not deterministic, but they are highly influential in the way they shape individuals’ prospects and decision making. Moreover, this approach allows us to understand how institutions affect the relationship between groups, not just the impact on the state as a whole. While many policy practitioners and academics have debated whether federalism, decentralization, and communal autonomy have made states more or less likely, this dissertation proposes that the true impact of those institutional innovations is on the group level, not the state level, and may have vastly divergent impacts on the degree to which ethnic groups are likely to respond to ethnic electoral appeals. Take, for instance, two of the most common aspects of consociational institutional design: proportional representation and territorial autonomy. Proportional representation is usually justified on the grounds that it lowers barriers to entry for smaller groups, facilitating their entry into the institutions of political decision making. In terms of the theory advanced here, proportional representation does nothing to change the degree of commonality of interest, and has little impact on the electoral viability of majority ethnic groups—as the majority, they can already be reasonably assured of a victory under virtually any electoral system. But it drastically increases the electoral viability for small minority groups. Switching from a majoritarian electoral system to a PR system will therefore increase the likelihood of ethnic voting for minority groups

16 but do nothing among majority groups by altering the electoral viability dimension. Territorial autonomy, on the other hand, operates on the dimension of interest. Taking a national-level minority and decentralizing them to become a majority group in a subnational unit drastically alters the voting calculations for voters. If groups are granted control over local schools, police forces, distribution of public funds, etc., then identity no longer serves as a meaningful basis of redistribution. Decentralization may actually decrease the commonality of interest of the minority group by increasing within-group competition, and lower the attractiveness of ethnic appeals. Viewed within the theoretical framework advanced here, it becomes clear that two of the most common policy prescriptions for divided societies–proportional electoral systems and decentralized governance–may actually push voters in opposite directions. Disaggregating the effects of these institutions along the two dimensions proposed here helps to resolve one of the recurring frustrations of applied institutional design: that institutional change is a blunt instrument, and the results of changes are often difficult to predict without disaggregating all the ways they alter incentives. The framework proposed here may also help explain why ethnic voting is so resilient in so much of the world. Frustrated international observers have noted that patterns of ethnic voting persist despite obviously damaging consequences. Blatantly corrupt, incompetent, and— in some countries—even criminal candidates win the support of constituents, for whom the only possible qualification can be coethnicity. Experimental studies which seek to provide voters with new information on the complete incompetence of their candidates has met with mixed and limited results, falling on deaf ears in many circumstances. The theory outlined in this dissertation suggests that while there is an informational component to ethnicity, ultimately ethnic voting persists because it is likely to deliver greater benefits to voters than mobilization along some other social cleavage. If this is the case, then new information about candidate quality or policy consequences is unlikely to sway voters to some other political option. Changes in demographic status or in the institutional environment in which elections take place, however, are more likely to alter voters’ and politicians’ calculation, and could result in drastic changes in voting behavior, even in the short-term. As such changes are relatively rare at the national level in most countries, the enduring persistence of ethnic voting should hardly be surprising.

17 By understanding ethnic identity as a resource which may be more or less valuable, and then disaggregating that resource into its two constituent parts, this dissertation seeks to avoid two of the biggest methodological pitfalls in studying ethnicity and identity politics: observational equivalency and spurious correlation. Since every human being has traits that could by some deﬁnition be referred to as “ethnic,” ethnicity is omnipresent in all forms of political action. A correlation between ethnic identity and voting could be due to any number of factors. Eth- nicity is highly correlated with other factors that could be related to political interest. Ethnic groups often cluster geographically, suggesting that they may share policy preferences caused by regional interests. Ethnicity may also correspond to other hereditary traits which may be passed down through generations such as class, professional training, property, or position in social hierarchy. Ethnic groups may share common cultural or religious practices, and protecting those practices and ensuring their continuation may require protection or support from the state and society. In any of these cases, the link between ethnicity and voting behavior would be entirely epiphenomenal, driven by issues that are closely related to ethnicity, but not actually a result of ethnicity per se. By establishing research designs that give voters the option of voting for ethnic parties and for parties that deﬁne themselves along other cleavages, this dissertation seeks to tease out when ethnicity drives voting behavior, and when it is spuriously correlated with something else. Moreover, the theory provides a way to reconcile competing notions of nationalism and voting behavior that don’t always line up the same way in majority and minority communities. Throughout the developed world, ethno-nationalist voting has been on the rise. Perhaps the most prominent example has been the more frequent expressions of white nationalism in the United States accompanying the election of Donald Trump to the Presidency, but such sentiments have also been seen among major European parties. Much of the literature on ethnic politics has focused on the vulnerabilities of minority ethnic groups, the threats to democratic legitimacy engendered by longstanding grievances in minority communities, and the potential for destabilizing secessionist movements led along ethno-cultural lines.12 Explanations of ethnic political mobilization based on grievances among a dominated minority community make it

12See the seminal contribution of Gurr (1970) and the voluminous research produced by the Minorities at Risk project (Asal, Pate & Wilkenfeld 2008).

18 harder to explain why ethnic voting may be increasing among numerically dominated majority groups. After all, how can it be that a numerically dominant group could be disadvantaged in a democracy under a system of majority rules? The theory proposed here suggests that because majority groups always have the greatest likelihood of winning in electoral contests, they may, all else equal, have a greater incentive to support ethnic appeals made by politicians than minority group members. As societies diversify—as has been the case in both Western Europe and the United States—it is in the majority group where we should expect to see increased ethnic voting. This is because as minority groups become larger, the benefits of ethnic representation to majority groups increases. Implementing policies that redistribute resources from minority groups to majority groups now increases the pay-off to those voters in the majority group. The theory here helps explain why “ethnic grievances” may counterintuitively be more effective tools of electoral mobilization, and therefore more prominent, among ethnic majority groups who already dominate most aspects of society. Lastly, the approach outlined in this dissertation seeks to incorporate human agency into the relationship between identity and voting behavior. To paraphrase Marx, people may make their own identities, but they do not make them as they please. It has long been acknowledged that identity repertoires in the individual can be expansive, with identities that overlap, contra- dict, or operate independent of each other. Proving an identity exists, even that an identity is strong within a given community, is not always sufficient to explain why that identity takes on political salience. This dissertation argues that when identities become politically salient, it is the result of choices. These choices are made by both voters and politicians in response to the circumstances in which they find themselves. Much of the literature has argued that ethnicity is a type of social cleavage around which political actors can mobilize, and that ethnicity adds to a menu of potential options for political mobilization along with class, region, etc. This line of reasoning has led to a fruitful research agenda, but largely ignores what it is about ethnicity that becomes so important in many places in the world. This “black box” is especially perplexing given that ethnic identities are relatively arbitrary and by their nature would seem to have little relevance to politics and the organization of the state. This dissertation approaches this problem by analyzing what it is, exactly, that voters might expect to get when they vote their ethnic iden-

19 tity. It starts from the premise that when voters respond to an appeal for votes made on ethnic grounds, they are eﬀectively declaring that they have chosen ethnicity as the best possible of all the avenues available to them. Examining when voters respond to ethnic appeals over other appeals more fully appreciates the way that voters behave in the real world, and may provide some insight into how we can expect electoral democracies to manage relationships between groups.

1.4 Plan of this Dissertation

This dissertation starts by outlining a theory of ethnic voting, which explains when the value of ethnic representation is at its highest. This theory makes predictions of divergent behavior between ethnic majority groups and minority groups even within the same country, which I then test through an analysis of two ethnically divided societies: Bosnia and Herzegovina and Latvia. In the Bosnian case, an analysis of split-ticket voting within a single electoral contest shows that individual voters respond to ethnic cues entirely differently depending on demographic circumstances. In the Latvian case, an analysis of voting results over time shows that politicians respond to the incentives created by this individual-level behavior by emphasizing ethnic identities only in some circumstances, subject to majority or minority status. The two cases in combination are used to support the main argument of this dissertation: that group identities can provide different benefits to different populations, and that these incentives are crucial to understanding how ethnic divisions affect voting behavior and elections. To examine when ethnic appeals are most successful, this dissertation employs a mixed- methods approach in both countries. First, I use ethnographic methods to identify exactly how parties and candidates use ethnic identification when campaigning to voters. First examining the historical roots of the ethnic divisions in both countries, I then identify how political entrepreneurs employ ethnic rhetoric, historical allusion, and divisive symbols in order to convey their own ethnic identities and their intentions to favor coethnics if elected. This is used to produce a qualitative data set, which is then mapped to quantitative voting data to facilitate an analysis of when ethnic parties are likely to succeed, and when they are likely to fail.

20 This dissertation is organized as follows. The next chapter develops the theoretical discussion begun here in greater detail. Chapter 3 discusses the empirical strategy and the design of the research project, explaining the mixed-methods approach employed in testing the theory. I then turn to the Bosnian case to demonstrate individual-level reactions to the incentives described by the theory. Chapter 4 is an ethnographic description of ethnic politics in Bosnia, outlining how ethnic labels are used in contemporary Bosnian politics. It also describes the data set which is used in the primarily quantitative analysis of Bosnian voting behavior presented in Chapter 5. In Chapter 6, I turn to Latvia, discussing both the history of Latvian interethnic relations, and providing an overview of how Latvian politicians rely on ethnic messaging in their electoral campaigns. The analysis presented in Chapter 7 shows that Latvian politicians adapt over time as predicted, tailoring their political messages diﬀerently depending on whether their message is likely to appeal to majority or minority group members. Chapter 8 provides concluding remarks.

21 2

A Theory of Group Size and Ethnic Voting

In this chapter, I outline a theory of how incentives for supporting ethnic parties can vary with the interaction of group demographic status and electoral institutions. I argue that ethnic voting is most likely among those voters whose ethnic identities are able to fulfill two equally important functions. The first is that ethnic representation must provide an outcome that voters are likely to desire. An ethnic group collectively wielding the power of the state must be able to pass or administer state policies in a way that provides some advantage to group members, making ethnic representation worthwhile. The second is that the ethnic identity must be instrumentally useful in solving coordination dilemmas and mobilizing voters in order to gain access to the institutions of power. Organizing along ethnic lines must be a reliable path to winning elections, making ethnic representation a viable possibility in the minds of voters. This chapter makes four important points. The first is that these two dimensions for benefiting and coordinating ethnic group members vary independent of each other. A voter may find that her ethnic identity is very closely correlated to policies or state actions that could benefit her directly, but that her ethnic group has virtually no chance of electing group representatives. It is equally possible for a voter to find themselves in the position of easily electing an ethnic champion to office, but also see that ethnic representation would be of little use. The second point is that these two dimensions are directly influenced by group size, but in different ways. All else equal, the larger a group is, the higher degree of usefulness the group identity has in winning elections, but the lower the degree of benefits to doing so. Thirdly, because of the way

22 in which group size influences these two dimensions, majority and minority groups will have fundamentally different assessments of what voting for an ethnic party gets them, and their willingness to vote for ethnic parties will diverge. Finally, political institutions dramatically alter these calculations by setting the political boundaries which establish majority and minority status, and the rules under which the groups must interact. Understanding ethnic voting requires an appreciation of the differences between majority and minority groups, and the ways that institutions can mitigate or exacerbate these tendencies. This chapter proceeds as follows. First, I define the terms used in this dissertation. Secondly, I outline differ from other bases of citizen-party linkages. Third, I outline a theory of how these two separate dimensions—the benefits of ethnic representation and the ease of winning office after mobilizing along ethnic lines—vary with group size and produce different strategic calculations between majority and minority groups. In section four, I outline the observable implications of the theory, and testable predictions of ethnic voting derived from this theoretical outline. In the fifth section, I give examples of the theory in practice, using visual depictions, and examples drawn from real-life voting behavior. The final section concludes.

2.1 Deﬁnitions

2.1.1 Identity

For the purposes of this dissertation, an identity is any category that can be used to classify or describe an individual. This definition is not universally accepted throughout the literature. Under this definition, identity is a discrete and bounded unit, rather than an all-encompassing abstract idea. Indi- viduals do not have a single “identity” which guides their actions or sense of self, but instead are composed of an enormous catalog of identities. Following Chandra (2012b) and Fearon (1999b), identity here is categorical.1 These categories include the politically relevant identities that form the mainstay of identity politics and cultural studies (e.g., “immigrant,” “gay,” “Evangelical Christian,” etc.), but they also include other categories that define one’s role in

1See also Barth (1969), Laitin (1998).

23 society and day-to-day activities, such as professional identities (“lawyer,” “teacher,” “doctor”) and interpersonal identities (“mother,” “husband”) which help categorize individuals into different groups based on societal expectations. Identities can also stem from values orientations or preferences over the way things should be (“libertarian,” “vegan”). Identities even include the trivial (“morning person,” or “Mets fan”). Under this definition, an identity is any discrete category that can be used to describe a person. In this way, identities are the tools used by human minds to resolve the problem of establishing the boundaries between self and others, and determine the proper relations between individuals and the social world in which they operate.2 This definition differs from others in both academic and popular usage, in that no identity is assumed to be of primary intrinsic importance, and makes no a priori distinction about the relevance of ethnic identity. This is somewhat at odds with a recent trend in academic scholarship of paying closer attention to the social relevance of particular identity categories.3 This research has fruitfully expanded on our understanding of the ways in which identity categories structure human behavior, but is not appropriate for this study. The goal of this dissertation is to explain the value of ethnic identities to voters and politicians over other identities like partisanship, ideology, or class. To make assumptions or decisions about which identities are most politically relevant risks tautology. The definition used here more fully appreciates the fact that individuals often have complicated identity repertoires which may be more or less relevant to their political preferences and decision-making. This definition has the added benefit of allowing for individuals to be composed not of a sin- gular “identity,” but a plurality of “identities,” which more closely tracks with the ways most individuals view themselves. People usually understand themselves through a repertoire of broad categories that apply to other individuals in society and create communities of similarity: terms like “graduate student,” “social scientist,” “American,” “midwesterner,” “man,” “caucasian,” “foodie,” etc. These categories are not mutually exclusive, and many of them operate within nested relationships. For example, “midwesterner,” as used here is a subset of “American.”

2For psychological interpretations on the problem of identifying self and the relation between self and society, see Cooley (1922), Mead (1934), Jenkins (1996), Jenkins (2000), Barth (1969) and Tajfel (1982). 3See Posner (2004a) who makes the case that ethnic cleavages should be assessed in light of their relevance to historical and social context within countries, or Cederman, Wimmer & Min (2010) who argue that research should take into account whether or not an identity category is represented by a political organization.

24 The “American” identity reflects commonality with other Americans, whereas “midwestern” differentiates between Americans. Employing a definition that allows for individuals to be composed by the combination and interaction of multiple identities also takes into consideration that the combination of identities can themselves be strong identities. Drawing on set theoretic terms, Crenshaw’s (1991) concept of intersectionality and the fruitful literature that it inspired4 provides a strong reason to believe not only that individuals view themselves in terms of multiple identity categories, but that the interactive combination of those identities can have strong correlations to people’s life experience. While intersectional scholars have often highlighted the ways in which intersectional identities can reinforce social discrimination and exploitative systems of privilege and oppression, the same definitions of identities have been employed by other scholars to identify ways in which societal diversity may be managed by democratic institutions. Stepan, Linz & Yadav (2011) stress the importance of “multiple but complementary sociocultural identities,” around which democratic states can form. In their analysis, state institutions and policies can either stabilize or antagonize inter-group relations based on the identity categories they privilege. The most stabilizing identity must usually be built around the state itself, not ethnic, linguistic, or religious identities. Using their example to show how nested identities are important to understanding political outcomes, Canada can emerge as a successful robustly multinational society by privileging the civic identity of “Canadian” which is universal throughout the country, over “Anglophone” or “Francophone,” which is more divisive.5 While the definition used in this dissertation does not preclude strong links between identity and circumstances, or between identity and behavior, it is not required by the definition itself. In this way, the conceptualization of identity differs strongly from other influential studies in the field. White (1992, p. 6) defines identity as “...any source of action not explicable from biophysical regularities, and to which observers can attribute meaning....” Under this definition, identities only exist if they are somehow linked to action. It is not possible to have an irrelevant identity, since identity only exists according to this definition if it is the source of an observable

4For discussion of how positivist political science can beneﬁt from methodologies based on the intersection of discrete identity categories and intersectional approaches, see Hancock (2007), and Dhamoon (2011). 5On the distinction between civic and ethnic identities, see Brubaker et al. (1999).

25 action. It is also not possible to have an identity to which individuals attach enormous personal importance, but little social or political importance. The definition employed here operates independent not only of action, but also of sentiment. Wimmer (2013) differs from my definition in stressing that identity communities must be not only categorical, but also behavioral, based on some notion of homophily, or preference for in-group member welfare over out-group member welfare. Clearly this is the case in many identity communities around the world, as evidence by the global ubiquity of ethnic favoritism, racial discrimination, and strong patterns of endogamy. But in the definition used here, preferences for like group members in social interactions are not a necessary requisite of identity. An identity may be accompanied by strong preferences for interactions with in-group members (as is frequently the case with ethnic, religious, linguistic, or cultural identities), or not (as is the case with professional, interest, and physical-attribute identities), and still be considered an identity all the same. My definition differs in that identity is purely descriptive, and can exist independent of action of emotional attachment. In fact, most of the identities which belong to an individual are mean- ingless for studies of social phenomenon and political behavior. Few political scientists would be interested in identity categories of “left-handed,” “allergic to peanuts,” or “able to drive a stick shift,” but these would nonetheless count as identities under my definition. A tolerance for irrelevance is necessary for a study such as this. To understand when an identity becomes politically relevant, we must employ a definition that does not assume relevance at the outset.

2.1.2 Ethnicity

Deﬁnition of Ethnicity

In this dissertation, ethnicity is defined as an identity based on attributes associated with descent. Under this definition, which borrows from Chandra (2004), “ethnicity” is a subset of “identity.” All ethnicities are identities, but not all identities are ethnicities. Ethnicity is but one of a very large number of identity categories that an individual can have, and is in no way more or less important to a person’s behavior, values, or political preferences than any other identity. But because ethnic identities are defined by descent, they are not readily changeable by any human agency. Whereas many identity categories the result of human choices, ethnicity is largely

26 the result of history and circumstance surrounding an individual’s birth—one can make choices that lead to becoming an “accountant,” a “college graduate,” or a “marathon runner,” but one is born “Sicilian,” “Arab,” or “Japanese.” This is somewhat at odds with the traditional usage of “ethnic” and “ethnicity,” which in common English parlance dates back to the middle ages. As Hutchinson & Smith (1996) point out, for most of the twentieth century in the United States, the term “ethnics” referred to Catholic, Jewish, or Slavic peoples who were recent immigrants to the US relative to the longer-established Anglo-British community. This is similar to the medieval usage of “ethnic,” which reflected the Biblical Greek usage of the term ethnos to refer to non-Abrahamic pagans. The definition here tries to clarify the concept by removing from it any sort of meaning derived from normative value attachments or inter-group power dynamics. Under this definition the British-descended Americans had no less of an ethnic identity than more recent immigrant ar- rivals. Likewise, the definition seeks to strip any prerequisite for a specific type of behavior and employ a minimum standard. This differs from Handelman (1977), which requires regular interaction with in-group members; Schermerhorn (1978, p. 12), who requires that ethnic groups have a similar culture and define themselves through it; or Smith (1986, ch. 2), who requires a sense of solidarity and an attachment to a geographic homeland. These are all very common phenomenon associated with ethnic identities, but the definition of “ethnicity” employed here any of them as a matter of definition. This minimalist definition is more appropriate for the present study, as it allows us to examine the link between ethnicity and political behavior without assuming that such a link exists ab initio.

Deﬁnition of Ethnic Voting and Ethnic Party

An ethnic party is defined as a party which claims to represent people of a certain ethnicity. Ethnic voting is defined as voting for an ethnic party. Under this definition, ethnic voting is something that an individual does regardless of the actions of other group members. This differs from other studies of ethnic voting, which rely on the degree to which an ethnic group votes together as a bloc.6 This approach has produced an impor-

6See Bratton, Mattes & Gyimah-Boadi (2005), Dunning & Harrison (2010), and Ishiyama (2012) for recent examples from Africa; Birnir (2007), Teney, Jacobs, Rea & Delwit (2010), and Heath, Fisher, Sanders & Sobolewska

27 tant and useful literature analyzing the variation in contexts where ethnicity is a strong predictor of vote choice. Nevertheless, the approach imposes some strong scope conditions, and allows for studies only in those contexts where ethnicity is already assumed to be a salient political cleavage on which political parties could mobilize voters. Such an approach may overstate the degree to which it is the ethnic cleavage driving voting behavior, instead of geographic or institutional factors. In countries where ethnic groups are divided into different geographical areas, different ethnic groups may find themselves living in extremely homogeneous electoral districts with equally homogeneous choices on their ballots despite living in a very diverse country. The fact that many voters are supporting coethnics could be the result of a calculated decision on the part of voters in response to their own political preferences and the options available, or it could be epiphenomenal to ethnicity entirely. In the United States, all people from North Dakota vote for other people from North Dakota. This is not because “North Dakotan” is the most salient individual identity, but because in the US electoral system it is impossible to vote for a con- gressman who is not from one’s own state. Looking only at voting cohesion as an indicator of ethnic voting risks attributing importance to identities which may be epiphenomenal. The definition of ethnic party employed in this dissertation relies instead on a party claims to do, and how it defines itself. This also differs from several prominent studies, notably Horowitz (1985), who defines ethnic parties as parties which derive most of their support from a single ethnic group. My definition is closer to Kitschelt (2001) and Chandra (2011) which are based on a party’s self-identification with a particular ethnic community. There are several benefits to this approach. Definitions of ethnic party which rely on voter support base effectively preclude the possibility of a “failed” ethnic party—all of the ethnic parties under Horowitz’s definition are inherently successful because they are defined by already having supporters. This definition is better suited for a study of the effects of successful ethnic parties versus other types of political parties, but ill equipped to examine why some ethnic parties succeed in winning office and others do not. In a similar vein, defining parties as ethnic based on how they present themselves to voters not only allows for an incorporation of human agency into a discussion of ethnic voting, but disaggregates the agency of voters and of political elites. Basing an analysis on definitions

(2011) from Europe; and Huber & Suryanarayan (2016) for South Asia.

28 which allow for parties to propose ethnically deﬁned platforms, but nevertheless fail to mobilize voters and win support based on that platform, allows us to better appreciate the individual factors and the mechanism by which ethnic identities inﬂuence vote choice.

2.2 Linking Ethnicity and Voting

Under the deﬁnitions above, people are composed of multiple overlapping and complimentary identities, with ethnicity comprising only a small fraction of an enormous network of identities that make up an individual. Why then is it that in so many contexts, appeals to ethnic identities dominate over other appeals to other identities? I argue that ethnic identities are useful to both politicians and voters in ways that other identities are not. Elections pose several challenges for both voters and politicians. Candidates want to get elected. Voters want to elect candidates who will enact favorable policies. Both sides are con- fronted with informational asymmetries and scarcities, as well as coordination problems. In certain cases, ethnic identities can help resolve these issues, helping a candidate win her contest, and a voter obtain a favorable outcome. Ethnic political parties are most likely to do well when ethnic identities are most useful in resolving these problems. In this section, I argue that ethnic appeals are likely to be made successfully in those contexts where they hold the greatest capacity to resolve these problems.

2.2.1 Ethnicity and the Challenges of Elections

On election day, voters are presented with a list of candidates, and can choose who to support. This process is not a simple one. To make an informed decision, voters first need to know what candidates and parties are will implement what policies, and how each plans to govern. Then, they need to figure out how to cast their votes in a way that is most likely to bring about the best possible outcome. Even in countries with consolidated democracies, effective impartial media, and developed political parties, the processes of information-gathering and decision-making can be difficult. In developing or recently democratized countries, these challenges can be even greater, given the high costs of accurate information and the difficulty in predicting electoral

29 outcomes in rapidly-changing political contexts. Candidates for office face similar obstacles. First, they need to develop proposals and campaign messages that will persuade voters to support their candidacy and elect them to office. Given the difficulty in assessing voter preferences this is not always easy. Moreover, they need to communicate these proposals and positions to voters in effective and believable ways. For the same reasons voters have trouble obtaining information, candidates have trouble disseminating it. Even if candidates have the luxury of developed party outreach capabilities or media outlets they can use to broadcast their positions, voters may not be in a position to receive or believe that information. In all of these circumstances, ethnicity can be a useful resource for both candidates and voters. Given the relative fixity of ethnic identities, and the ways in which ethnicity is easily observable, ethnic identities can be employed by candidates to persuade voters, and by voters to choose leaders. Obtaining information on voting decisions is difficult (Lupia, McCubbins, Arthur et al. 1998). In multiparty elections, voters may be faced with a large number of candidates or parties, with few resources to differentiate them. Especially in new democracies or in situations of high electoral volatility, voters may have no reliable information available on how a specific party would run the country if elected. When candidates are not well known, voters cannot rely on reputation or past performance. Where parties are newly formed, their ideological positions may not be firmly established (Tavits 2008). These problems are not unique to new democracies, though, as even in advanced post-industrial states assessing the consequences of specific policies is complicated. Moreover, most voters have weak incentives to invest the time and resources needed to educate themselves on their voting decisions (Downs 1957, Riker & Ordeshook 1968, Campbell, Converse, Miller & Stokes 1980[1960], Popkin & Popkin 1994, Martinelli 2006). Even further compounding the issues is that candidates in almost all contexts may have strong incentives to misrepresent their own interests, hiding potentially corrupt motives behind rhetoric of public service. Filtering the competent from the corrupt is difficult to do without observing politician behavior, which requires not only further investment from voters in obtaining information, but also clarity of responsibility in retrospective assessments (Manin, Przeworski & Stokes 1999, Maravall 2005, Tavits 2007, Chang, Golden & Hill 2010). Because

30 of these challenges, it is not always easy for voters to decide who they want in office. Voters must not only decide which parties or candidates would make the best leaders, but also consider social choice and collective action challenges produced by the way votes are counted and translated into political representation (McKelvey & Ordeshook 1972, Cox 1984). Electoral outcomes are the result of the choices of all citizens as well, meaning that an individual voter needs to think not only about how her own vote choice affects the outcome, but how it is likely to affect the outcome given the choices of all other candidates. One of the most significant implications of this is the possibility of vote-wasting. For voters concerned about ultimate outcomes, voting for a most-preferred candidate may not always the best course of action. Voting for a second-choice candidate when a first-choice is not viable may actually be more likely to tip the election in favor of that candidate and prevent a less-desired outcome. The risk level depends on the preferences and calculation of other voters which are not always immediately obvious, and the risk of coordination failure in voting can be quite high, and voters’ best possible course of action may actually be to vote for a party whose stated position is not necessarily closest to their true preference. If a voter’s most preferred candidate or party has absolutely no hope of winning an election, but their second choice is a viable candidate in a tight race, that voter may increase their likelihood of casting the decisive vote in favor of a more preferred candidate by voting strategically. These are all complicated decisions, and make voting a potentially onerous process for voters. Political parties and candidates who want to win elections can therefore better their chances by helping to resolve these issues for voters. The best strategy is to convince voters both that electing them is a positive outcome that will benefit the voters, and that they will not be wasting their vote by doing so. While voters get to choose who they will support, parties and candidates get to decide what platform they will present to voters. Their objective is to propose a platform and cultivate an image that voters will find appealing. They can emphasize economic issues, stress cultural issues, link themselves to a specific charismatic leader, etc. To win, this platform must have a broad enough appeal that it can persuade a group of voters large enough to carry the party to victory. Even after deciding on their positions, parties need to communicate this information

31 to voters. In nearly all modern political systems, the number of citizens is large enough that candidates cannot interact with each potential voter individually, and so must rely on mass media or the party outreach capabilities. This imposes costs and difficulties on the parties and the candidates, especially in developing or recently democratized countries where mass media and party infrastructure may be underdeveloped. Ethnicity may be helpful in resolving some of these informational and communicational problems between voters and leaders. First of all, ethnic identities can be used to convey a commonality between voters and politicians, and therefore presumably common policy interest. For voters who struggle to differentiate one would-be leader from another in the absence of reliable information, this may be a helpful heuristic. Absent any other information on candidate preferences, priorities, or competences, voting for a coethnic may be a good option since shared ethnicity creates the possibility of shared interest through the mechanism of common descent (Ferree 2006, Birnir 2007). For voters wanting to elect leaders who share their priorities, ethnicity may be a useful for proxy for how a candidate is likely to behave once elected. Secondly, information on ethnic identities can be transmitted more credibly and efficiently than other candidate characteristics. Ethnic markers are often present even in a candidate’s name, and signals conveying ethnic information can be displayed by the candidate in speech, mannerism, dress, or cultural practice. In this way, it can be much easier for a candidate to convey to voters their ethnic identity than it is to provide a nuanced understanding of their policy position. The fixity of ethnic identities also helps overcome potential moral hazard problems caused by informational asymmetry. Because ethnicity is fixed and unchanging, voters need not worry about “bait and switch” tactics, where candidates may promise one policy, but enact another once in office. Once a candidate has established her ethnic identity to the voters, voters need not be concerned that the candidate will change when elected. A commitment to help coethnics is fairly credible, since the candidate would herself benefit from such policies. In other words, it is impossible to “flip-flop” on a position based on one’s ethnic identity, since such an identity cannot change easily. The visibility of ethnicity also helps to predict how other people may vote, helping to resolve collective action problems and coordination dilemmas. Since ethnic identities are just as

32 highly visible among voters as they are among candidates, politicians can also easily observe potential supporters’ ethnicities. Politicians can therefore use demography as a proxy for how well-received an appeal to a specific ethnic group is likely to be. An appeal to support a specific group and provide them with preferential treatment may not be universally acceptable among the beneficiaries themselves (as their sincere preference may align more closely with class, ideological, cultural, or some other cleavage), but it is guaranteed that no one outside the group will find this argument persuasive. In other words, an appeal to favor group A at the expense of group B may not be acceptable to all A’s, but it will never be acceptable to anyone in group B. Politicians can therefore assess the ceiling on the support for a specific ethnocentric position by gauging the population ratios of their would-be constituencies. For this reason Chandra (2004) argues that ethnic parties succeed on the basis of “ethnic head counts”—making ethnic appeals to voters only in constituencies only where the ethnic group is large enough to win an election. Ethnic distinctions can therefore convey valuable information to voters, and distinguish between insiders and outsiders—all reasons why ethnic identities might be useful as the basis of linkages between voters and candidates. But just because ethnic identities provide some value in the electoral process does not mean that people will always choose to vote for candidates exclusively on the basis of ethnicity. It is entirely possible that voters or candidates could have reasons to disapprove of the limited information provided by common ethnicity, determining they have better prospects by mobilizing along some other cleavage or identity. Voters may decide that ethnic labels are actually not useful at all in conveying commonality, or may have access to more accurate information on candidate quality or positions. They may believe that their ethnic identity has nothing to do with their political interests, and find coethnicity unper- suasive grounds for voting. It is even possible that individuals believe the representation of their own ethnic group in the organs of political power could be disadvantageous to themselves, and prefer not to have a coethnic in power. If voters do not see a link between their identities and their political preferences—in other words, if how they want the state to govern is completely unrelated to their ethnicity—then ethnicity is not a resource useful in pursuit of interest. Like- wise, if politicians see that voters are not persuaded by ethnic appeals they have no reason to campaign on them. This could also be the case if there are simply not enough voters to mobilize

33 to win office. For those politicians, ethnicity is not useful to winning office, and of little value. Politically speaking, some ethnic identities are more useful than others. Not all ethnic groups will find it a good idea to mobilize along ethnic lines because such mobilization simply does nothing for them. This may be because the best-case scenario of ethnic electoral mobilization— the take-over of the state by ethnic champions—would do nothing to advance the interests of the individuals involved. It may also be the case that ethnic mobilization is not likely to be successful, resulting in coordination failure and vote wasting. But for some groups, ethnic mobilization is likely to lead to success, and provide a substantial policy pay-off. Seizing state power in democratic elections through electoral mobilization is both possible and efficacious. Acknowledging the various degree of usefulness of ethnic mobilization is key to understanding when groups mobilize along ethnic lines and when they don’t. If ethnic mobilization is likely to result in benefits for voters and office for ethnic elites, then everyone has an incentive to politicize ethnic cleavages. Ethnic appeals will win out over non-ethnic appeals, and group ethnic representation is most likely. If ethnic mobilization is unlikely to result in benefits for voters and office for ethnic elites, then the ethnic identity has no value, and non-ethnic appeals will come to dominate. Understanding when voters are most likely to support an ethnic appeal and when candidates are most likely to make them requires us to think systematically about when ethnicity is a useful mobilizing tool and when it is not. I argue that it is important to distinguish between two separate functions for which ethnic identities can be instrumentally helpful: setting policies advantageous to individual voters, and mobilizing groups of voters large enough to win elections. While the literature has often conflated these two, they are separate dimensions, and can vary independently of each other. Ethnic voting is most likely for those voters who see both conditions met.

2.2.2 The Beneﬁts of Ethnic Representation

What, speciﬁcally, do voters have to gain from electing leaders who explicitly identify as representatives of an ethnic group? Literature across the social sciences suggests that there are three separate strands of beneﬁts to be gained by ethnic political representation: intrinsic psycholog-

34 ical rewards, the provision of club goods through policy enactment, and patronage. All three of these benefits are directly correlated with demography, and thus contribute to the differences between majority and minority groups. The smaller the group, the more likely they are to benefit from electing ethnic champions. Minority group members therefore stand to benefit more from ethnic representation than majorities. Perhaps the most straightforward benefit is the emotional benefit that comes from seeing a coethnic in office. As outlined by social identity theory (Tajfel 1978), and self-categorization theory (Turner, Hogg, Oakes, Reicher & Wetherell 1987), individuals understand their own role in the world and the value they offer to society through their membership in larger groups. This link is partially instrumental, but it is also emotional, as individuals may derive their own sense of self-esteem and self-worth from the overall perceived prestige and worthiness of the group to which they belong. In psychology, this affective commitment—or emotional attachment to groups—has been shown to produce a willingness to incur individual costs in order to maintain the strength of the group (Allen & Meyer 1990). Political representation is one of the most prestigious honors a group can hold, and the power that comes with holding elected office can produce a strong sense of pride. For Horowitz (1985) this sense of self-worth is especially important in ethnically divided societies where divisions may exist between “backwards” and “advanced” groups. A lack of representation for one group in state power may raise concerns of disenfranchisement, and discrimination; but it is also personally disheartening, as group members feel that their group exclusion reflects a lack of their own worthiness or capability. In voting for a party that campaigns on a platform of representing a single ethnic group, voters may be seeking the validation that comes from empowering one of their own, accessing the “psychic benefits” associated with group representation. But politics is not all symbolic, and voting is not all emotional. Modern states do important things, implementing policies that have enormous impact on society and the material well-being of individuals. It may therefore be the case that the benefits of ethnic representation can come from common interests shared within ethnic communities. Since ethnicity is by definition associated with descent, it is correlated with other individual characteristics that are also associated with descent. Since most people speak the same language as their parents, linguistic divides

35 often correspond with descent, and therefore may also be highly correlated with ethnic identity. The same is true with other social behaviors and ideas that are passed down from one generation to the next, including religious beliefs and cultural practices. Voting for a party which claims to represent a specific ethnic group therefore means voting for someone who has a similar background and presumably similar preferences with regards to these issues. In modern, industrial- ized economies, states also have strong incentives to foster linguistic and cultural homogeneity throughout their territories (Weber 1976, Gellner 1983). Which language is chosen as the official language of the state determines which group of people must incur the additional costs of learning new languages and those who have the privilege of speaking their native tongue. In addition to the loss of prestige and cultural validation associated with speaking a subaltern language, those who do not incur these costs and develop their language skills risk social exclusion, political disenfranchisement, and economic deprivation (Csergő 2007, Liu 2015, Stepan 2015). Ethnic representation is therefore an instrumental tool to avoid the homogenizing policies that would force assimilation and adaptation on a group. Ethnic representation can also provide direct material rewards to voters in the form of patronage. Fearon (1999a) argues that ethnicity is an extremely useful basis for the distribution of political patronage and clientelistic benefits. Since ethnic markers are not easily changed, they eliminate the common problem of monitoring in clientelistic democracies, and facilitate the identification and maintenance of minimum winning coalitions. In campaigning on promises of representing a specific ethnic group, especially in the so-called “patronage democracies” where the state primarily redistributes services and benefits to political supporters in exchange for votes (Chandra 2004), parties can imply that voters should support them by virtue of their ethnic identity because they will favor coethnics once in office. The motivation to support an ethnic party is not based on a cultural precept or any idea about what it means to be a member of a certain group, it is simply about favoring coethnics at the expense of other groups. Ethnicity is valuable because it serves as a reliable basis for differentiating “insiders” and “outsiders”—or winners from losers. The only real policy to speak of is to redistribute away from one group and towards another at every possible opportunity. There are therefore three different genres of benefits that could be had by electing an ethnic

36 champion to office: psychological, policy, and redistributional. Which benefits voters value most, or why a politician might focus on offering one over the other is an important question, but beyond the scope of this dissertation. But all three share one important commonality: the value to an individual voter of each type of benefit is higher for small groups than it is for large groups. All of these benefits increase as group size decreases. All else equal, a small group will receive more of each of the three different types of benefits described than a large one. On psychological benefits, there are two explanations why this is the case: one rooted in human nature, the other in mathematics. Psychologically speaking, smaller groups are more likely to hold greater meaning and import for their members than larger groups. Brewer (1991) has shown that those identities which become the strongest are those which bond individuals to others in the same category while at the same time distinguishing the self from others. The “ideal” identity in this regard—i.e., the one to which an individual is most likely to develop the strongest personal attachment and sense of worth—is one that differentiates an individual from the largest possible group of people while associating with a small but still viable community. This is closely related to the mathematical relationship between group size and electoral prospects. If, for example, political leaders were chosen at random instead of through elections, then the odds of a person from a specific ethnic group being chosen would be the same as their proportion of the population. If a single group represents 90% of the population, then that group has a 90% chance of having one of their own members chosen. This makes being elected a much less likely event—and therefore much more meaningful—for groups which represent a very small segment of the population. As an example, note the historically important moments when members of numerically small (and usually historically disadvantaged) groups become head of state: Barack Obama as the first black President of the United States, K. R. Narayanan as the first Dalit President of India, or Leo Varadkar as the first LGBT Taoiseach of Ireland. The elections of these leaders are often seen as pivotal or inspirational moments, i.e., those that provide the greatest psychic benefits. On policy matters, small groups also benefit more from ethnic representation than large groups. Smaller groups have more to lose, and are more likely to be targeted, while larger groups are more likely to dominate the political and economic sectors. At best, a simple pre-

37 disposition to employ coethnics due to social networks or physical proximity could put smaller ethnic groups at a serious disadvantage. At worst, ethnocentrism or racism could encourage open discrimination and hostility between ethnic groups. In a situation with fewer potential targets and more potential threats, smaller ethnic groups are more likely to suffer than larger. If there is a homogenizing pressure in society, as is frequently associated with industrialization, globalization, and other forms of economic development, the languages and cultural practices which are most likely to dominate are those of the largest ethnic groups. A party intending to protect ethnic group interests can actually do a lot to alleviate these concerns if elected to office.7 They can implement state policies designed to protect minority languages from assimilationist trends.8 Ethnic parties can also implement anti-discrimination policies, legally protecting their ethnic constituency or ensuring their access to state jobs and resources, giving the community the increased strength of state assistance to level the playing field in relation to larger groups and prevent assimilation. Smaller groups also benefit more from between-group redistribution than larger groups, even if only by mathematics. As an over-simplified example, imagine a situation where everyone pays a certain percentage of their income into a general government fund, and the government then redistributes resources on the basis of identity. If the resources are redistributed only to a single group, then the payoff to each individual is larger if the group is smaller: the pie is simply divided into fewer pieces. The most lucrative between-group redistribution would be when a very small group is able to tax a very large group and distribute resources to its members. A situation where majorities are taxed to pay minorities is relatively rare in truly democratic governments on any large scale.9 In democratic regimes, the more likely outcome is the dominance

7See Gurr (2000) for a theoretical overview and speciﬁc case studies of minority communities’ elevated risk of assimilation, and discrimination. 8See Flores (2008) for documentation of such programs in Latin America. In Bolivia, for example, the Ministry of Education subsidizes newspaper supplements in the Aymara and Quechua indigenous languages spoken by 14.6% and 21.2% of the country, respectively. Whereas such a policy is hardly necessary for Spanish-language media, state intervention produces media that these ethnic communities would most likely not have access to if left to the free market given the relatively small demand in light of the size and the poverty of the intended audience. 9Since majority rule is generally the guiding principle of democratic governance, this type of extremely regressive redistribution is usually only associated with authoritarian regimes, such as the government of Saddam Hussein which disproportionately favored Sunni Muslims (only 30% of the country) at the expense of majority Shi’a, or the regressive apartheid regime of South Africa, under which white citizens—never more than 20% of the population—controlled the bulk of state resources. Such extremely regressive group-based policies can result in massive windfalls for beneﬁciaries, sustained by the repression of everyone else.

38 of a minimum winning coalition (Riker 1962, Fearon 1999a, Posner 2004b) which redistributes resources to its own members. Each additional member beyond that required to win elections ultimately reduces the pay-offs to each individual member, a result of the logic that smaller groups benefit more than larger ones. Regardless of the specific type of benefit voters gain by ethnic representation, it is always more beneficial for smaller groups to be represented on the basis of ethnicity than larger groups. Voters from smaller groups derive a greater sense of validation and psychic benefits from seeing coethnics in office than larger groups. Those voters in smaller groups also are more likely to benefit from public policy goods like linguistic and cultural protection than larger groups, since larger groups are more easily able to defend their communal interests and practices on their own without state intervention. And voters in smaller groups are able to gain more from excludable targeted redistribution than larger groups, since the net benefit per voter is larger by virtue of dividing resources among a smaller number of citizens. All of this suggests that all else equal, minority ethnic group voters will potentially have more to gain by electing ethnic representatives than majority group voters.

2.2.3 Ethnicity as Coordination Device

Usually, the winner in an election is the one who got the most votes. As a system of majority rule, democracy favors larger groups, and bigger political factions win out over smaller ones. Throughout history the identities which are usually most important are those with substantial numbers of people, e.g., large-scale societal classifications like “protestants,” “workers” or “land owners.” Small groups are generally irrelevant, as political viability is linked to group size in a democracy. Size is a necessary but not sufficient condition for electoral victory. A group must also be at least minimally coordinated. Voting is fraught with risks of coordination failure. Voters may, among other things, “waste” their vote by supporting a losing candidate when voting for a second-choice candidate may be more likely to change the outcome of the election and produce a more desired outcome. Formal models of electoral outcomes have shown that where the probability of being the deciding vote in favor of a second-choice vote is sufficiently large, the

39 probability of being the deciding vote in favor of a first choice candidate is sufficiently small, and the difference in utility between electing a first choice and a second choice candidate is sufficiently narrow, then voters have a strong incentive to strategically vote for a second-place candidate (McKelvey & Ordeshook 1972). Scholars of strategic voting have pointed out that this type of collective action can be fraught with difficulties, and that not all groups may be equally equipped to resolve them (Aldrich 1993, Cox 1997). Voters need to ascertain the likelihood of their preferred candidate winning. If this is low, then they must also ascertain which options are most viable, in order to identify how to cast a strategic ballot. Ethnicity can help with this, because ethnic attributes are often highly visible and generally unchanging. Gauging the viability of ethnic group representation is easier than measuring the potential popularity of policy positions. Whereas it is not particularly easy to assess whether someone supports a more ideologically moderate or extremist candidate without polling them directly, the willingness of someone to support an appeal to a specific ethnic group is more readily obvious. An appeal to use the power of the state to the benefit of a specific group will not be very popular with voters who do not belong to that group. Demographics set the upper limit to how far an appeal on the basis of ethnicity can go: an appeal to group solidarity and superiority may not resonate with all in-group members, but it is clear that it will not resonate with any out-group members. All else equal, the coordinative capacity of ethnicity in voting is highest for large groups. In a democracy, the larger number of voters almost always wins. Increasing group size results in a larger group of voters who may respond to the appeal, increasing the electoral viability of the ethnic group. Larger groups are more likely to be perceived as strong electoral contenders than smaller groups. If, for example, a voter sincerely prefers to have her ethnic identity explicitly represented in parliament, and she is a member of an overwhelming majority, then she knows that this group winning office is a plausible outcome. She may not know exactly how many of her coethnic voters also prefer ethnic representation instead of representation along class, ideology, or some other social cleavage, but she knows that there are at least enough voters who could respond to an ethnic appeal to carry the group into office. A voter with similar preferences in the minority group, however, faces a different calculation. The total number of voters likely

40 to respond to an ethnic appeal is ultimately capped by the size of the group. If that group is very small, then the likelihood of an ethnic party being politically viable is much lower. It is a mistake to argue that big groups win and small groups lose every time in a democracy, and it is inaccurate to say that all minority groups face insurmountable obstacles in every situation. The relationship between a group’s size and its political viability varies greatly with the institutional context in which the group finds itself. As a voluminous and fruitful literature on institutional design has shown,10 electoral rules have an enormous influence on which groups are and are not politically viable. The biggest determinant of group size is district magnitude. Duverger’s law famously holds that majoritarian, first-past the post systems tend to result in two-party systems (Duverger 1962). Under a single-member district plurality voting rule, only a single candidate can be elected from each district, so candidates with small bases of support are completely unable to win election. Mathematically, the system usually favors the largest vote getter disproportionately, since a candidate will always receive 100% of the seats (i.e., one), and therefore 100% of the political authority, despite receiving only a majority or in some cases a plurality of the vote. This tendency changes the strategic incentives of both voters and elites. Elites, mindful of the fact that only the largest faction in a district can win any political power, will tend to concentrate into either the dominant party or a single opposition, since doing so maximizes the chances of election. Voters, in turn, will likely limit the range of voting choices to the top two parties or candidates, aware that voting for a small faction is most likely to result in a “wasted” vote with little to no chance of swaying the election. Under proportional electoral systems, on the other hand, the calculations are quite different. Since legislative seats—and therefore, political power—are awarded in proportion to the votes received, smaller parties have a chance to win power in proportion to their support they receive among voters. Despite this fact, the differences in group size impact the risk of calculation are important regardless of the electoral system. Even in the most permissive and proportional of electoral contexts, larger groups have an easier time gaining access to office than smaller groups. In a highly proportional system with a very low threshold, the risk for both large and small groups may be so small as to be more or less inconsequential, but we can nevertheless say that there

10See Duverger (1954), Rae (1967), Sartori (1976), Taagepera & Shugart (1989), Lijphart (1994), and Stoll (2013).

41 is a difference between the groups. It is possible, for example, for a group comprising 60% of the population of an electorate to split into 12 equal-sized groups, and still come in above a 5% electoral threshold. A minority population that is only 10% of the population can only split into two. In other words, a voter who wants ethnic representation in a majority group requires a lower degree of uniformity of preferences among her coethnics than a comparable voter in a minority group. The decision-making calculus of voters when deciding to support an ethnic party thus consists of two interactive components: an assessment of the benefits that would accrue to them under a system of ethnic representation, and an assessment of how likely such representation is to obtain. The two dynamics vary not only independently of each other, but with opposite relationships to group size. The benefits of ethnic representation increase as group size decreases, whereas the ease of installing ethnic parties in power increases as group size increases. These dynamics suggest that the decision of whether or not to support an ethnic party plays out very different between majority and minority ethnic groups, simply by virtue of their differences in their relative share of the population.

2.3 Institutions and Group Status

For ethnic voting to be most appealing, ethnic parties must offer benefits that voters want, and it must be feasible for them to win elections. Because these two explanatory concepts are directly correlated with group size, majority groups and minority groups face different incentives to support ethnic parties. Majority groups have an easier time electing ethnic parties, but relatively little to gain from the ethnic representation they provide. Ethnic minority groups would have much to gain through ethnic representation, but face much greater difficulties in electing ethnic parties. But “majority” and “minority” are relative terms. Whether one is a majority or minority group member depends not only on absolute group size, but also on the borders drawn around ethnic communities by administrative and electoral institutions. This interaction between demography and institutional context can alter the relationship between majority and minority groups, and dramatically change the calculations of individual voters.

42 In many contexts, the administrative unit which enacts policies and the unit which elects leaders are not the same: policies are usually enacted by the state, and leaders are chosen by electoral districts. The distinction matters, because in almost all countries, institutions and demographics interact to ensure that the ethnic make-up in the state as a whole and in the electoral district are drastically different. In a perfectly unitary country composed of a single electoral district, state and electoral district are the same. Yet practically speaking, very few countries fall into this category.11 In many countries, ethnic groups which find themselves as state-level minority groups are geographically concentrated so that they constitute majorities within specific areas (Lublin 2014). Where electoral districts are delineated geographically, small groups that are disproportionately concentrated within districts will have a much higher ability to clear electoral hurdles than those groups which are dispersed equally across electoral districts. This may be the case due to concentration in some kind of historic homeland—as is the case with the Welsh in the UK— or urban migration patterns—as is the case with black voters in US cities like Detroit.12 In these cases, national-level minority groups make up an overwhelming majority in the districts in which they live. These groups may thus be guaranteed representation through ethnic voting since they easily exceed all practical barriers to entry. An ethnic group’s political viability can also be changed intentionally for ethnic groups by institutional designers hoping to empower minorities or facilitate group representation. In many countries, electoral districts are drawn around historically relevant regional boundaries, which often coincide with ethnic groups’ traditional homelands. The border between Flanders and Wallonia is written into Belgian electoral law, dividing the country into districts where Dutch-speakers make up the majority and French-speakers make up the majority, with relatively few bilingual districts. This ensures that most voters—regardless of ethnic identification—find

11To my knowledge, the only countries that meet this perfect overlap in lower house elections are Israel, Moldova, The Netherlands, Paraguay, Serbia, Slovakia, and Timor-Leste, and many of those use diﬀerent electoral systems for upper house or presidential elections. Kyrgyzstan is composed of a single electoral district, but also contains special seats reserved for ethnic minority populations. It is therefore extremely rare for the demographic characteristics of electoral districts to perfectly match the demographics of the state as a whole. See Bormann & Golder (2013) and Bird (2014). 12The 2010 US census revealed the city of Detroit to be 84% African-American, while the state of Michigan as a whole was 79% white, making Detroit a majority black enclave in an majority white state. Illustrative of the ways in which local-level demographics can impact the viability of ethnic representation, all but one of Detroit’s mayors since 1974 have been African-American, while every governor in Michigan’s history has been white.

43 themselves living in a district where their ethnic group is a viable political bloc, despite French- speakers comprising a national-level minority. Nigerian electoral law requires that electoral districts be drawn in a way the respects “cultural affinity,” resulting in ethnic groups divided into districts where they each comprise a local majority. In both the Belgian and Nigerian cases, electoral laws are designed to give each ethnic group greater chances or representation and lower risk of coordination failure than they would have otherwise had in a majoritarian election where all groups voted together. In the most extreme cases, ethnic groups are guaranteed representation by explicit legal protections and mandates. Lebanon is the archetypal example of ethnic quotas in electoral politics: the constitution requires that president be a Maronite Christian, the Prime Minister a Sunni Muslim, and the Speaker of the Parliament a Shi’a Muslim. Other countries include explicit guarantees or ethnic representation, including New Zealand, which sets a minimum quota for ethnic Maori¯ candidates in parliament; and India, which provides guaranteed representation for a list of traditionally underrepresented caste and tribal groups. Bird (2014), drawing on Vukelic (2012), estimates that 28 countries currently have explicit legal protections guaranteeing representation for specific groups in the lower house of parliament, with many more relying on implicit systems like ethnic gerrymandering to ensure minority groups representation. These institutions are specifically designed to make minority representation possible. By guaranteeing a certain level of representation to a specific group, they effectively eliminate the barriers to entry for small groups. Whereas in a purely majoritarian election, a very small group has no chance of being elected, these institutional innovations segregate minority groups onto an ethnically-segregated ballot, essentially turning a national-level minority group into a majority group within their own electoral unit. This arrangement changes the viability of ethnic representation from being very low to very high. These institutional configurations change the incentives for voters to support ethnic parties. Voters must assess both the benefits they are likely to receive from electing ethnic parties to power, and the likelihood of victory for those parties. While both of these are directly related to demography, they change based on the group’s position within the institutional environment. The former is determined at the level at which political decisions are made, while the latter

44 Ease of Ac- Beneﬁts of Ethnic cess For Ethnic Representation Groups Group Size at Group Size at State Function of Electoral District Level Level Highest when Group is Small Group is Large “What will I get “How easily can Voter Calculus if my group is in my group win?” power?"

Table 2.1: Two Components of Ethnic Voting is determined at the level at which leaders are elected. Ethnic voting should be most likely among those voters who have high values on both dimensions. Ethnic minority groups are most likely to support ethnic parties when they are very small, but geographically concentrated within electoral districts where they are guaranteed to win elections if they mobilize around a single party or candidate. Ethnic majority group are most likely to support ethnic parties when they are faced with a large minority group that is geographically concentrated into their own ethnic enclaves. The general outline of the theoretical components determining ethnic voting are outlined on Table 2.1.

2.4 Visualizing the Theory

Incentives for ethnic voting therefore vary based on demographics and institutions, and majority and minority group voters face fundamentally diﬀerent incentives and strategy concerns when deciding whether or not to support an ethnic party. In this section, I present a more intuitively understood approach by showing how the dynamics outlined above would impact voters in more relatable situations. First, I illustrated this argument using the abstract example of an election in a very small, diverse society. I think show how the incentives to support an ethnic party vary based on group status using the real-world example of ethnic Albanian voting in a variety of contexts.

45 2.4.1 Ethnic Voting Decisions in a Very Small Polity

Imagine the polity as seen in Figure 2.1. It is ethnically divided between two ethnic groups, indicated here by the colors blue and yellow. Suppose there is an election for a single legislator, and three parties are running. One is an ethnic party from the yellows. It advocates for yellow supremacy, supports the yellows to the exclusion of the blues, and proposes that the state should show preferential treatment towards yellows. Another party is an ethnic party from the blues. It takes the exact same positions as the yellow ethnic party, but in favor of its own coethnic blues to the exclusion of yellows. The ﬁnal party is non-ethnic. It believes that ethnicity should not be the basis of state policy, and instead advocates for some other basis of legitimacy. Obviously, this over-simpliﬁed example ignores the enormous complexity of real-life party systems and elections. “Non-ethnicity” is not a party platform or proposal as much as it is the absence of one, and thus would probably not exist as an actual campaign platform in the real world. Nevertheless, in this extremely limited example we can already say something about likely outcomes and voting behavior using the framework outlined above. Take for example,

our voters labeled as and in the Figure. We know that is a yellow, and that yellows are the majority group in this polity. We can immediately see that she will not support the blue supremacist party. That party, if elected, would relegate her to second-class citizen status, rejecting her identity and culture, and taxing her to fund programs that would disproportionately beneﬁt the blues. Her choice is therefore between the non-ethnic party, and the ethnic party

representing her own group. The inverse is true for . He is a blue, and therefore has no incentive to support the yellow ethnic party for the same reasons. Let us examine the beneﬁts of ethnic representation to both of these people. From this it is clear that has much to gain from the election of the blue ethnic party. He is in a minority that

2 comprises only 7 of the population, and thus his group’s victory would be extremely prestigious. He would also be the recipient of a fairly large windfall if the state decided to tax yellows to redistribute to blues, since 7 people would be paying in, but only 2 people would be collecting

the beneﬁts. The rewards to from the election of the yellow ethnic party is much more modest. Her group outnumbers the blues more than two-to-one, so even if the non-ethnic party is elected there’s a good chance that her group will be represented in the institutions of power by

Figure 2.1: Viable Yellow Party; Non-Viable Blue Party

Figure 2.2: Two Electoral Districts Increases Viability for Blue Party

Figure 2.3: Greater Diversity Increases Beneﬁts from Yellow Party

47 numbers alone. A redistribution policy taxing blues to pay yellows reaps an extremely modest benefit: with 2 citizens paying in but 5 citizens collecting, the spoils are spread too thin to really count for much of anything. Admittedly, without knowing anything about the citizens beyond their ethnicity, or other facts about the parties running, we cannot say for certain whether these benefits would be more or less preferable to our two voters than what the non-ethnic party has to offer. We can nevertheless say that the benefits of ethnic representation to are higher than they are for . The ease of access to office, though, is exactly the opposite, in that it is much higher for

than for . While may have the most to gain from ethnic representation, it is entirely impossible for him to achieve. Even if his fellow blue coethnic supports the blue ethnic party, he knows for certain that no one in the ethnic majority will support the minority ethnic group’s party, and since they outnumber him it is impossible for the blue ethnic party to win. Even if the yellow voting bloc splits their vote perfectly between their ethnic party and the non-ethnic party, one of those parties is guaranteed to have at least three votes, outnumbering the maximum two votes that the blues can muster. In this instance, has a very strong incentive to support the non-ethnic party. The two blue votes supporting the non-ethnic party may be enough to tip the scales in favor of the non-ethnic party over the yellow ethnic party, which ultimately would be a good thing for both blue group voters, avoiding ethnic appropriation from the yellows. For

, voting for an ethnic party is a dangerous choice that is ultimately detrimental to his own interests.

, however, has the luxury of voting for whoever she truly prefers. Either the yellow ethnic party or the non-ethnic party is guaranteed to win this election, and even without knowing the true preferences of her coethnics, her chances of casting the pivotal vote in this extremely small polity are quite high. She need not worry about accidentally allowing the blue ethnic party to win, since this outcome is impossible in this situation, so is free to vote for either party. Again, we cannot say exactly which party she should vote for without going into more speciﬁc and idiosyncratic details about her situation, her preferences, or the nature of the competing parties. But we can still compare the relative group-level incentives in this scenario. Voters in the minority group have zero incentive to support an ethnic party, whereas voters in the majority

48 have a moderate incentive to support ethnic parties. Ethnic voting in this situation is therefore more likely among the members of the majority than the minority. Now examine the scenario in Figure 2.2. Instead of electing a single legislator, the polity is going to elect a parliament of two legislators, from two separate electoral districts. The population of the country is the same, but the institutional rules governing the election have changed. Since the demographics of the entire polity are the same, the beneﬁts of ethnic representation

are the same as they were before: very high for our minority blue voter , but more moderate for our majority yellow voter . But the likelihood of each ethnic group winning has changed. For , the situation is almost the same. She now ﬁnds herself in an ethnically homogeneous district. She knows that no one in her district will vote for the blue ethnic party, and so is free to vote her true preference just as she was before. But for , the game has changed substantially. Since the minority blues are now a district-level majority, the blue ethnic party has suddenly become viable in a way that it wasn’t before. If his fellow coethnic votes for the ethnic party, then it is possible for the ethnic blue party to take the district. The likelihood of victory has gone from impossible to relatively high, causing a drastic increase in his willingness to support the ethnic blue party. The change to the institutional rules has done nothing to alter his motivation for supporting an ethnic party, but it has made the viability much greater, and overall increased his willingness to vote for the ethnic party. In fact, in this case, ’s incentives to support an ethnic party may be higher than ’s, since ethnic representation is now viable for both voters, but so much more lucrative for the minority group member than the majority group member.

Note that the institutional change has not altered ’s incentives, but it has altered ’s. Even though and share an ethnic identity, their incentives are the same in Figure 2.1, but divergent in Figure 2.2. In Figure 2.2, now lives in an ethnic enclave of blues. He remains a state-level majority group member, but is now a district-level minority. Even if he has a true preference for the yellow ethnic party, he knows that neither of the other two citizens of the district will support that party, making victory impossible.13 essentially ﬁnds himself in the same position as in Figure 2.1. Even if he truly prefers a yellow ethnic party candidate, such an outcome is

13Technically, a three-way tie is possible in this scenario. In order to retain parsimony, the implications of this are ignored, as such an outcome is likely only in the extremely over-simpliﬁed example presented here. The important insight for the theory from this example is that in the majority-blue district, a blue ethnic party victory is possible, as is a non-ethnic party victory, but a yellow ethnic party victory is not.

49 impossible due to the high barriers of entry in that district, and so his best strategy is to support the non-ethnic party in hopes of casting a pivotal vote if the two blue citizens split the vote between non-ethnic and blue ethnic. Let us make one more change, this time to demographics, as reﬂected in Figure 2.3. Here, the two-district legislature is still in eﬀect, but neither district is purely homogeneous. The

4 yellows are still a majority, but their dominance has declined somewhat, in that they are now 7 5 of the population instead of 7 . For , this demographic change only increases the likelihood of ethnic voting. She is still a majority of the polity as a whole, she is still a majority in her district, but by swapping one of her coethnics for an ethnic outsider, ethnic redistribution has become more lucrative. There is now one more ethnic outsider whose resources could be appropriated, and one less coethnic to share spoils with, increasing her individual gains. The two groups are also now approaching parity, and yellow’s status may not be as guaranteed as it once was. This demographic change has only increased her own willingness to support an ethnic party by changing the benefits of ethnic representation, while maintaining ethnic representation as a viable outcome. In this overly-simplistic example it is not possible to say definitively who will support an ethnic party and who will not. But it is possible to gauge the relative attractiveness of ethnic voting in each case. Each individual voter’s incentives to support an ethnic party vary due to their demographic circumstances within both their district and the polity as a whole. Because of these incentives, even voters of the same ethnic group can find themselves in varying circumstances based on their local demographics. Majority and minority group status can dramatically change the predictions of support for an ethnic party.

2.4.2 Ethnic Voting in Reality: Albanians and the Albanian Diaspora

The examples using the abstract figures above show how a voter’s relative preference for being represented by an ethnic party can change based on the demographic circumstances in which they find themselves. This abstract example ignores much of the complex realities of social and political life in the real world. But a theory of ethnic voting that appreciates both the potential pay-offs of ethnic representation and the ease of wining if mobilizing along ethnic lines also

50 helps to understand the variation in support for ethnic parties in much of the world today. Take, by way of example, the ethnic Albanian community. The greatest concentration of ethnic Al- banians is, unsurprisingly, in Albania, where they constitute roughly 98% of the population. A close second is Kosovo, where they comprise 90% of the population. A large ethnic Albanian diaspora spreads throughout Eastern and Central Europe, with substantial ethnic Albanian minority communities residing in North Macedonia (25% of the population), and Montenegro (5% of the population), as well as smaller numbers in Italy, Romania, Turkey, Greece, and Croatia, with overseas enclaves in the United States and Canada. Albanian ethnic and national identity is extremely strong, both within Albania and Kosovo, and throughout the international Albanian diaspora.14 Yet independent of the strength of that identity, the incentives to vote for an ethnic party vary widely depending on context. Imagine an Albanian voter living in Tirana, the capital of Albania proper. Here she finds herself part of an almost completely homogeneous population. There is hardly any minority group to speak of in the country. All leaders are virtually guaranteed to be ethnic Albanians by mathematics. In this situation, successful ethnic mobilization is guaranteed to lead to political representation. The population is almost entirely Albanian, so it is entirely possible for the group to clear any formal or informal threshold to winning electoral office if mobilized successfully. But the benefits of ethnic representation are very low in this case. Albanian interests are hegemonic; all candidates and voters are likely to share the same ethnic identity, so they share the same ethnic interests. There is no large group of ethnic outsiders from whom voters may need protection, nor is there an outgroup to target for the reallocation of resources. Non-ethnic distinctions are much more relevant in this case, since those will likely be much less uniform throughout the country, and a voter would likely have much more to gain by redistributing from rich to poor for example, or from lowland areas to highland areas, than they would on ethnic labels. The theory predicts that these relatively low levels of benefits to be had through ethnic voting will lead to some other identity or social cleavage being chosen as the basis of political competition, as voters are mobilized by class, ideology, or region. This is, in fact, what has happened in Albania, as the primary political cleavage in recent elections has been between

14See Koinova (2013) and the Special Issue: Albanian Migration and New Transnationalisms (2003) in Journal of Ethnic and Migration Studies.

51 Socialists and Liberal Democrats. Regional and linguistic identities are somewhat strong predictors of voting behaviors in Albania, but neither party explicitly campaigns on a doctrine of ethnic identification, or a promise to benefit ethnic Albanians within the state specifically. Now, consider an ethnic Albanian voter living 150 miles away in Pristina, Kosovo. Kosovo is still overwhelmingly ethnic Albanian, as Albanians represent 90% of the population, but there is now a sizable group of ethnic minorities. Unlike Albania, Kosovo spent much of the twentieth century as a part of Yugoslavia, and the first several years of the twenty-first century as a part of Serbia, and many of its ethnic minority citizens are Serbs and other ethnic Slavs. Serbs are guaranteed a minimum level of political representation, and ethnic Serb voters are segregated into a separate ballot to cast votes for their own leaders.15 In this situation, ethnic Albanian parties still face no immediately obvious barriers to winning an election. Albanians are still an overwhelming majority of the country, and any politician who mobilized even a fraction of the ethnic Albanian majority would be ensured to gain some representation in parliament. The difference is that the benefits of ethnic representation have gone up. As the minority community here is larger, ethnic Albanians face a slightly larger threat from ethnic outsiders. Ethnic minorities are also an extremely sensitive issue in Kosovo, given the still-recent history of violent conflict during an ethnic civil war in the late 1990’s between Serbs and Albanians. International factors exacerbate this domestic tension, as the Republic of Serbia maintains territorial claims on Kosovo, refuses to acknowledge its declaration of independence, and advocates internationally for the cause of ethnic Serbs within Kosovo. Many voters may therefore be more susceptible to a political platform which promises to neutralize threats— real or perceived—to the majority Albanians by ethnic minority groups. In fact, despite their common ethnic identity, a typical voter in Kosovo and a typical voter in Albania will likely show drastic differences in their support for ethnic parties. In a study of Albanian political parties and elections, Barbullushi (2016) shows that ethnic and ethnonational appeals to voters are far more common in Kosovo than

15Some studies suggest that Serbs are also more likely to suﬀer economic deprivation than Albanians in Kosovo, putting them in the position of a vulnerable minority (Bhaumik, Gang & Yun 2006). While the Serbs are not the focus of this example, the theory does also help understand Serb voting behavior in Kosovo. As a vulnerable minority group representing a small segment of the population, they have much to gain through ethnic representation. As the beneﬁciaries of guaranteed representation, they also have very low barriers to entry. The theory therefore predicts high ethnic voting, which is exactly what happens. In both the 2014 and 2017 elections in Kosovo, nine out of ten reserved Serb seats were won by the Serb List party, an ethnic Serb interest party.

52 they are in Albania. Reflecting the international environment, parties in Kosovo are much more likely to see sovereignty for ethnic Albanian people as a central component of their platform. It should be noted that a direct comparison between Albania and Kosovo in this regard may not be appropriate, as it is virtually impossible to disentangle domestic ethnic issues from international ones in the Kosovar context, but not in the Albanian: Albania has a much longer history of independence and sovereignty, whereas Kosovo was a former Yugoslav province. Nevertheless, the comparison illustrates how shared ethnicity does not always correspond to shared incentives to support an ethnic party. Now consider the situation in Tetovo, 60 miles from Pristina across the North Macedonian border. Here, an Albanian voter lives in a city that is almost three-quarters Albanian. Just as in Albania or Kosovo, Albanian is the main language heard on the street, and the Albanian double- headed eagle is seen far more frequently than the North Macedonian sun. Our voter is still part of a majority group within her electoral district—one of six in North Macedonia—with Tetovo falling in the same region as other ethnic Albanian enclaves like Bogovinje, Debar, Gostivar, Kičevo, and Vrapčište—all of which are majority Albanian. From the language spoken in the city it is obvious that Albanians are a local majority, and there is little concern that voting for a party that explicitly identifies as a champion of ethnic interests is a wasted vote. Nevertheless, the voter knows that she is in North Macedonia, a majority ethnic Macedonian country, and that there are substantial numbers of ethnic Macedonian voters who will probably not support a party claiming to represent the ethnic Albanian community. In this environment, the risk of vote wasting, or supporting an ethnic candidate who cannot win has changed little from the voters in Pristina and Tirana. What has changed dramatically is the benefits the voter is likely to receive from ethnic representation. Turning on the TV, she sees that most channels are broadcast in Macedonian—a Slavic language completely unintelligible from Albanian. She sees that while government services at the local level may be available in Albanian, services provided by the central government are more often in a foreign language. All but one of the state universities teach courses only in Macedonian and English, meaning that she must now learn another language if she wants to receive a higher education. Our voter now has serious, tangible grievances that result from her position as an ethnic minority. When a political party promises to represent

53 ethnic Albanian issues, it is clear what those are: language rights, access to state services, and representation in potentially hostile environment dominated by ethnic outsiders. If a candidate promises to deliver to ethnic Albanians their fair share of the country’s wealth, our voter now knows that redistributing resources from ethnic Macedonians to ethnic Albanians could result in some very real financial benefits for her and her fellow Albanians. Ethnicity may even be the dominant political issue in her mind when casting her vote, far more important than policy issues concerning class or other identities. This expectation is borne out by the facts in North Macedonia. The political divide between ethnic Albanians and ethnic Macedonians is extremely pronounced. Cross-ethnic voting is rare, as ethnic Albanian candidates campaign exclusively to ethnic Albanian audiences, while ethnic Macedonian candidates largely ignore Albanian voters. There is intense intra-ethnic conflict within the Albanian community, but parties mainly compete over who can best represent Albanian interests in opposition to the dominant ethnic Macedonians, making exclusively ethnic appeals to potential constituents. Now consider an ethnic Albanian voter thousands of miles away in the United States, where she finds herself in one of the very few Albanian enclaves in New York, Boston, or Los An- geles. Here she is unlikely to hear Albanian if she leaves her immediate neighborhood, and finding Albanian language media or cultural activities is no small task, if possible at all. While government services may be available in languages other than English, it is more likely to be in Spanish or Chinese than Albanian, making it clear to her that not only is she a minority, but she is one of the smallest and most demographically inconsequential minority communities. The benefits of having ethnic representation in this situation are even higher than for the voter in Macedonia. Since there is very little ethnic Albanian community to speak of, she can rely less on those in her immediate neighborhood for assistance and support. If a politician were to promise benefits directly to ethnic Albanians in the area, she would likely stand to benefit enormously, since few rivals would compete for benefits. Since there are so few ethnic Alba- nians in the area, setting aside even a fraction of one percent of government budgets to ethnic Albanian concerns would likely represent a huge windfall benefiting Albanians voter directly. A program designed to redistribute money from all tax payers to ethnic Albanians, say in the form of an ethnic Albanian scholarship fund, or Albanian homebuyers loan program, would have an

54 enormous funding base of non-Albanians to draw on and very few ethnic Albanians eligible. In this the potential benefits of ethnic representation are the highest of all scenarios reviewed here. However, any politician campaigning on ethnic Albanian issues has such a small voter base to draw from, that their campaign is likely doomed to failure if they campaign exclusively on Albanian interests. Excluding voters of other ethnicities to appeal to only ethnic Albanians is political suicide, since such a proposal is hardly likely to attract support from more than a very small number of voters. The barriers to entry to office in this context are insurmountable, as the proportion of the population that would support these appeals is so small compared to other ethnic groups.16 In all of the situations outlined above the degree to which Albanian identity is linked to political interests has changed because of differences in the demographic circumstances in which voters find themselves. Incentives to vote for an ethnic champion vary in each of the four examples because of changes in what she stands to gain from such representation, and the degree to which such representation is plausibly attainable. “Albanian” as a political identity has significance not because of the voters’ level of personal attachment, but because of the broader demographic and institutional context in which the individual voter finds themselves.

2.5 Conclusion

Elections are collective events where large numbers of individuals separately make decisions that determine what will happen to the group as a whole. Individual identities are clearly important to understanding those decisions, but it is no less important to acknowledge that the context in which those identities exist exert enormous inﬂuence over voting decisions. Voters do not think only of themselves and their own identities when voting, but also the diﬀerences between

16This is not just a hypothetical situation. In 2012, Mark Gjonaj, born in the Bronx to Montenegrin Albanian immigrants, ran for the New York State Assembly against incumbent Naomi Rivera. Touting his ties to the neighborhood where he grew up (and Rivera’s ongoing corruption investigations), Gjonaj won the election in part due to the support of Albanian-Americans living near Pelham Parkway and Allerton Avenue, one of North America’s densest Albanian diaspora enclaves. Gjonaj was not shy about his family’s immigrant roots, and was happy to court support from his coethnic neighbors, but never explicitly promised that benefits would flow exclusively to ethnic Albanians in his district; his official platform focused mostly on job creation and public health programs for senior citizens. Gjonaj was nevertheless invited to Tirana and awarded the Order of the Honor of the Nation—The Republic of Albania’s highest non-military honor—for being the first ethnic Albanian elected to public office in the United States. See Beekman (2012), Naomi Rivera Defeated Soundly in Assembly Primary By Hard-Working Challenger (2012) and Goldstein (2015).

55 themselves and others. They consider how their group position may constrain them or enable them, and then act accordingly. To understand how ethnic identities map onto preferences for ethnic political representation, we should acknowledge not just individual ethnic attachments, but also the relationships between groups, as influenced by relative group size. I have argued in this chapter that group size is an important variable in understanding the role of ethnic identities in voting decisions. This is not to say that group size is the only thing that matters. Clearly there are many other relevant factors that might influence a voter’s willingness to support an ethnic party. But group size is nevertheless important because it directly influences two separate dimensions which are central to making decisions on who to vote for: the potential policy gains of electing leaders who promise to champion ethnic causes, and the relative ease of actually installing such leaders in positions of power. Using this framework, I have argued that ethnic minority group members and ethnic majority group members have fundamentally different strategic calculations. Majority group members would easily win office if they efficiently mobilized their coethnic citizens, but would actually have relatively little to gain from doing so. Minority group members would stand to benefit enormously from their group controlling the state, but such an outcome is much more difficult because of the barriers to electoral victory for smaller groups. This argument suggests that ethnic voting among majority group members and ethnic voting among minority group members are actually different things, responding to different incentives and circumstances. As such, we should expect to observe different patterns of ethnic voting among majority and minority communities—a prediction I test empirically in Chapters 5 and 7.

56 3

Research Design

In the previous chapter, I outlined a theory of how ethnic voting may vary based on local demographic and institutional context. In this chapter, I outline the research design used in the rest of the dissertation to empirically test this theory. I survey the challenges the theory presents for empirical testing, and present a research design strategy intended to produce falsifiable predictions and test them. This dissertation relies on a mixed-methods strategy exploring two separate cases: Bosnia and Herzegovina, and Latvia. Qualitatively, I rely on ethnographic methods which survey the history of the ethnic divisions in these countries to identify the origins of the ethnic divides and the ways in which ethnic groups differentiate themselves from others in each case. Then, I explore how political actors use those differentiating aspects as part of their political strategies to attract supporters and win elections. From this, I generate a qualitative data set, coding candidates and political parties in both countries based on what type of appeal they make to voters, and whether they present themselves as representatives of a specific ethnic group or not. That data set is then matched to quantitative voting data for a statistical analysis to determine when ethnic appeals are successful and when they are not. These findings are tested against the predictions of the theory providing an opportunity for falsification. The two countries were chosen because of the opportunities they provide to examine voter- and elite-level implications of the theory. In Bosnia, the research design holds elite strategies constant, varying the institutional and demographic factors. Exploiting the fact that Bosnian voters may find themselves

57 in a majority group and in a minority group in a single election (depending on which level of government they are electing), I show that ethnic voting strategies systematically vary based on group demographic status. In Latvia, the research design exploits variation over time, holding institutions and demography constant. I show that political elites seeking office adapt their campaign strategies based on group status. Majority-group candidates become more likely to campaign under ethnic party labels, emphasizing their ethnic identities and appealing to voters on ethnic platforms. Minority-group candidates, on the other hand, become more likely to stress non-ethnic issues while campaigning. The first section of this chapter outlines the practical challenges to testing the theory outlined in Chapter 2. I also discuss the specific challenges of making causal claims based on a theory which relies primarily on structural, historical, and large-scale social context variables, and how I respond to those challenges. The second section explains how I respond to those concerns, and surveys the broad approach of this dissertation. It provides an outline of the empirical data collection and the mixed-methods approach employed in the empirical chapters that follow. This chapter is intended only to provide a broad overview, and to aide in interpreting the results of the dissertation and incorporating them into broader a literature which may rely on different methodological and epistemological approaches. More concrete details are also contained in the empirical chapters themselves.

3.1 Challenges and Approaches to Empirical Veriﬁcation

The main argument in this dissertation is that people are most likely to support ethnic parties when ethnic representation promises to produce outcomes that voters want, and when mobilizing coethnics will feasibly result in electoral victory. The dependent variable is ethnic voting; the independent variables are group size and electoral institutions. The mechanisms linking the independent variable and the dependent variables are the assessments of political actors regarding the benefits of electing an ethnic party, and the likelihood of being able to do so. In order to convincingly demonstrate the validity of this argument, it is necessary to generate falsifiable predictions that can serve to confirm or refute the theory. In this particular study there are sev-

58 eral potential pitfalls that must be addressed directly. The first is concept formation, ensuring that the theoretical concepts are clear, and that the empirical strategy coherently links observation to the concepts being studied. The second is reliable and valid measurement. Voting is a notoriously blunt instrument, wherein citizens synthesize many complicated issues and calculations into a single, observable action. Ethnic voting is characterized by voting for a party or candidate making a specific type of appeal. Measuring this is not always easy, and special care should be paid to ensuring that measures are valid. Finally, there is the issue of what kind of scientific claims can be made when the explanatory variables are largely structural. The independent variables in this study are the products of identities and institutions that in many cases are themselves the products of historical forces not readily altered by individual human agency. Research designs which rely on randomization and intervention to make strong causal claims are difficult in such a context. Shiveley (1980, p. 30) cautions that relying on everyday language can undermine rigorous scientific analysis, as daily usage frequently collapses multiple phenomena into a single concept. This is especially true of the dependent variable in this study, and previous studies have defined ethnic voting in a variety of ways. Ethnic voting could be understood as a vote for a party which claims to represent an ethnic group (Chandra 2011). It could also be observed when members of an ethnic group all vote together (Huber 2012), or when a party obtains all of its voting support from a single ethnic group (Horowitz 1985). In each of these conceptualizations, it is clear that there is an important relationship between ethnicity and voting, but they are nevertheless three distinct outcomes, possibly resulting from different causes and explained by different mechanisms. Likewise, the foundational concept of ethnicity is also defined in various ways in different studies. Some have defined ethnicity with regards to language, race, religion, birth- place, physiognomy, or other identity categories. These are all plausible conceptual foundations for ideas of ethnicity, but each definition would require a different measurement strategy, and different indicators of ethnic identity. Even with a theory based on cleanly defined concepts of the independent and dependent variables, measures on this topic are likely to be noisy. The theory outlined in Chapter 2 makes a series of predictions about the likelihood of majority group members and minority members

59 to support an ethnic party. It does not, however, claim to be an exhaustive theory of all voting behavior, as the claims are ultimately relative. It suggests that under certain conditions, ethnic voting is more likely than it would be otherwise. It does not claim that demographics and institutions are the only factors that matter in voting behavior. While I expect that the explanatory variables will produce observable differences between different types of groups, I do not discount the importance of ideology, class, candidate quality, and the many other important factors which could operate at the level of countries, parties, candidates, ethnic groups, and individual voters. As a result, observed voting behaviors are certain to vary for reasons not included in the main explanatory variables. Relegating these other factors to the error term of a statistical analysis could make precise estimations of correlation difficult. A research design which seeks to test the impact of demographics on ethnic voting should therefore acknowledge the importance of all other factors outside the theory, while nevertheless allowing for tests of the proposed mechanisms. One concern is that the independent variables are endogenous to an enormous number of potentially unobserved or unmeasured variables. Demographics are never randomized. Where people live, which ethnic group they identify with, and the size of their ethnic community is usually the result of complex historical factors. Since birthrates are correlated with economic status, the size of the group could in some instances be a proxy for economic variables which may actually be the cause of the observed correlation between demographic status and electoral results. Majority or minority group status is also usually the result of structural conditions or historical processes which may also have an impact on observed voting behavior. In the two cases analyzed here, this concern is especially pronounced, as ethnic diversity in Central and Eastern Europe is often the result of historical events and trends which are themselves worthy of serious study in their own right. In the former Soviet Union, the presence of minority group populations is the result of migration from the Russian metropole to periphery republics. Mem- bers of ethnic minority groups may therefore share certain traits, as they are the descendants of economic migrants who were settled in their current location by state policies designed to further the political aims of Soviet communism. In this case there is reason for concern about reverse causation. Electoral institutions which

60 shape relationship between groups are usually instated by political leaders who are in power. It could therefore be that ethnic voting puts in place a specific kind of political leader, who then implements a specific type of institution. If this is the case, then the voting behavior is the cause of the institutions, not the other way around. Demographic status may also be subject to reverse causation. If majority group members elect coethnic politicians on a platform of ethnic supremacy and coethnic favoritism, then those leaders may enact policies adversely affecting the minority population. If these policies encourage emigration, facilitate assimilation, reduce birthrates, or increase mortality, then the population of the minority group is likely to decline as a result of the electoral choices made by voters. Demography could therefore be the result of politics. As the present study uses observational data to test hypotheses, it is not possible to completely dismiss these concerns. The research design was structured as to hold constant the surrounding circumstances and unobserved variation as much as possible. In an experimental study, the dependent variables would be randomly assigned, with all unobserved causal factors equally distributed between treatment and control groups. Instead of aiming for a purely randomized controlled experiment, instead I focus on research designs which hold those other potentially unobserved causal factors constant. In Bosnia, I analyze split-ticket voting in the context of a single election and compare differences between voting at one level to voting at another. The long-term historical and structural variables are therefore held constant, since we observe voters casting two separate ballots, in most cases literally seconds apart from each other. In Latvia, I examine the same political candidates and their propensity to change platforms between elections. This design again holds the individuals constant, and operates on a short enough time table that the long-term structural factors are held close to constant between observations. This does not meet the “gold standard” of causal identification, but it nevertheless takes these extremely valid concerns over omitted variable bias seriously.1 These caveats should guide the interpretation of the findings. But it should also be noted that the relationship between the independent and dependent variables of this dissertation may not always be amenable to study using perfect causally-identified methods. Even given unlim-

1For a more in-depth discussion of the research design and its applicability to the two cases, see Chapters 5 and 7.

61 ited resources, a randomized controlled trial which could convincingly confirm or disprove the theory in a broadly generalizable way would be extremely difficult. Such a study would take a large group of people, randomly assign the social contexts in which they found themselves, and then observe the way they voted. This would be extraordinarily difficult, as group size and institutional rules are extremely difficult to manipulate in a laboratory setting. In real life, people know their ethnic identity and their relative group size because such things are communicated over time in day-to-day life, and credibly carrying these nuances and intuitions in test subjects is difficult in the artificial environment of a controlled laboratory. People understand their group’s position in society because of history and experience, and such intuitive understandings are not readily changeable in the short-term. But these things are important to study precisely because they are complicated, multi-faceted social elements. In this way, this dissertation responds to Huber’s (2017) concern that research driven exclusively by causal identification may bias the discipline away from studying questions “where the central explanatory variables are broad features of the social, political, or institutional context (Huber 2017, p. 104).” Ethnic identity may indeed be endogenous to political outcomes, which presents a potential problem for this study and others like it. But that endogenous relationship is also why it is crucial to study voting and social context together, since that relationship is so vitally important to normatively important outcomes like democratic representation and political stability. These questions are important to ask, and should not be ignored because establishing causal relationships is difficult. Rather, it falls to the researcher to be mindful of the risks of endogeneity as a challenge to convincing research designs. To better understand this tension, consider two separate “hard” sciences: medicine and astronomy. Medicine is fundamentally concerned with things that are smaller than humans. Its objects of inquiry are organs, biological processes, viruses, injuries, etc. These are all things that are either constitutive of human beings, or that interact with human beings. Astronomy, on the other hand, examines objects that are bigger than humans. Astronomers study the planet we live on, the forces which govern the motion of the stars, and the physical make-up of the universe. Driven by the needs of their subject, these disciplines have adopted drastically different approaches to advancing human knowledge. Political science as a discipline is situated in

62 between these two approaches. When we turn our attention to an individual’s ideology, preference, or behavior, we examine at things that are smaller than human beings. When we look at political parties, institutions, social groups, states, or international organizations, we examine things that are larger than human beings. Both medical and astronomical researchers have the luxury of studying things that are either smaller or larger than human beings. Political scientists, on the other hand, study both, and must learn from both of these approaches. Research designs driven by causal identification usually adopt the same methodological approach and epistemological position as medical researchers, wherein the gold standard is the double-blind randomized controlled trial. In the perfect RCT, the treatment is perfectly randomized within a population under study, such that subjects can be differentiated only by the presence of the treatment condition. As the only difference between the two groups is the treatment, any difference in average outcomes between the two groups can be causally attributed to the treatment. The double-blind nature of the study ensures objectivity by removing the immediate observer from the calculation entirely. Theoretically, since even the researcher performing the study does not know who is in which group, it is not possible to influence the outcome by intentional or unintentional interference. RCTs are important to medical research for good reason. First, they perform a crucial role by helping doctors and medical practitioners make vitally important treatment decisions on a day-to-day basis. These professionals need to know how likely a specific treatment is to have the desired effect, and whether that effect is likely to be superior to the effects of other potential treatments. Moreover, the double-blind RCT design helps eliminate biased results as a result of researcher subjectivity and potential conflict of interest. RCTs first gained in prominence in the United States in the 1950’s with the increase in mass production of pharmaceuticals intended for public consumption (Bothwell, Greene, Podolsky & Jones 2016). The biggest critique of RCT methodology at that time has since come to be seen as one of its strongest assets: it diminishes the importance of testimony and first-hand experience by experts and practitioners in favor of arbitrary and ostensibly universal findings which could be replicated by any researcher performing the same procedure under the same conditions. This feature is especially appealing in political science, where the subjects under inquiry are by definition, political. Even the most

63 disciplined and objective social science researcher is likely to have personal interests and preferences over many objects of study. RCT methods avoid such bias, as a properly executed RCT would eliminate the possibility of interference from researchers’ personal agendas and allow for more reliable claims to objective scientific facts. But RCTs by nature rely on intervention: the randomized administration of a specific treatment or not. In political science this is achievable in many contexts, but not others. Just like medical researchers, it is easiest when we study things that are smaller than human beings: individual-level decisions, information, or characteristics. But political scientists also care about variables that are not easily changed by direct human intervention. Societal-level characteristics, institutional environments, historically determined political realities, and slow-moving structural changes are also crucially important to political phenomenon, but are for the most part beyond the control of any single individual. Rather than ignoring these important variables which are less amenable to causally-identified research designs, we should instead adopt tools and approaches more akin to astronomy than medicine. Studying the motion of heavenly bodies and the basic physical properties of the universe is not something that can be done by intervention, since such an intervention is beyond human capacity. Moreover, even if humanity were capable of staging random treatments, doing so in many cases would defeat the purpose. As the goal of research in astronomy, cosmology, and astrophysics is to understand the way that the universe exists as is, intervention would be pointless. A counterfactual in the form of how the universe would exist if it were not our universe is a pointless question, since that universe would not be the real universe in which we live. In place of the RCT, the gold standard in astronomy is the perpetually developing relationship between theory and observation (Bradt 2004). Observations produce questions, and theories are developed to answer those questions. New falsifiable implications of those theories are then generated, and new observations are made that either confirm the implications and strengthen the theory, or reject them and require new theorizing. Progress is often made in astronomy through the development of new technologies that increase the observational capacity of researchers. The practice often requires careful consideration of the researcher’s own status as observer, and integration of that observational position into the research design itself. One of the most important discoveries of early modern astronomy—that the earth revolved around

64 the sun—was only made possible when theoreticians specifically articulated how the limited vantage point of the earth would influence the range of observations that were possible. The geocentric theory of the universe would be extremely easily disproved if the solar system could be observed from a great distance for a long period of time, but such observations are difficult even today. There is no experimental methodology requiring a treatment and control group that would facilitate this discussion. Instead, it was made possible through creative theorizing producing create falsifiable hypotheses, integrated with awareness of the limits of observational techniques. This is not to say that even within the two paradigmatic disciplines the two approaches are mutually exclusive. Astronomers conduct experiments which are useful in testing the micro- level implications of theories explaining macro-level phenomenon. A recent example or this would be the manipulation of particle motion to confirm of the existence of the Higgs boson as a way to strengthen the more general theory of the Standard Model of Particle Physics (CERN 2014). Similarly, medical researchers rely on observational studies of macro-level phenomenon to confirm theories of individual-level disease transmission, such as John Snow’s discovery that cholera was transmitted by waterborne pathogens based on observations of neighborhood-level spread (Freedman 2008). Ethnic voting can be studied advantageously using either approach, since there are both micro- and macro-level factors that are important to understanding the way ethnic identities impact voting decisions. For researchers most interested in persuasion, turnout, or preference formation, the medical paradigm is appropriate, and many scholars have fruitfully studied such things using RCTs. In this dissertation, however, I am most interested in how societal-level diversity shapes system-level outcomes. While individual-level decision-making is the mechanism that I argue links identity and electoral outcomes, in this study I am most interested in observing large-scale outcomes directly. As such, I adopt the perspective more akin to astronomy, and design a research approach which acknowledges the potential weaknesses and limitations in an observational approach, and compensates for those potential shortcomings by implementing strategies which minimizes potential erroneous conclusions.

65 3.2 Features of this Design

In order to ensure conceptual clarity and a clear match between concepts and measurement, this dissertation follows Chandra (2012b, chap. 2) and employs minimalist definitions of the subjects under study. In defining ethnicity as an identity based on attributes associated with descent, and ethnic voting as voting for a party which claims to represent an ethnic group, the definitions contain as few components as possible. These standards exclude extra characteristics which other scholars have sometimes employed, such as attachment to a specific homeland, common culture, or sense of shared fate. For this study, which rests on mechanisms linked to individual’s assessments, these extra factors could introduce other possibly mechanisms, and complicate hypothesis testing. Instead, these minimal definitions are derived from political actors’ descriptions of themselves and the world around them, producing a clearer match between the hypothesized mechanism, the concepts, and the measurement. If an identity attribute is understood to be associated with descent, then I accept it as an ethnic identity. If a party claims to represent a specific ethnic group, I code them as an ethnic party. By accepting the actors’ interpretations and claims of their own identity and political goals, I limit the need to impose my own standards when conducting the analysis. The empirical strategy to test the theory is fundamentally probabilistic. I argue that ethnic voting is likely for some groups in some contexts, but not others. This is a ceteris paribus claim that variation in demographic variables should be correlated with variation in ethnic voting holding all else equal. This is not to claim that ideology, culture, history, or individual preference do not matter. These are obviously important for voting and election outcomes. In using statistical methods which are non-deterministic, I differ from some other studies of political development and ethnic mobilization which see group-level characteristics and social context as important.2 In these studies, the ethnic-group or party is the unit of analysis, and hypotheses are tested using qualitative methods based on Boolean algebra. To produce statistically significant estimations which also allow for deviation from general tendencies, I rely on large-= datasets, with units of analysis much smaller than the group or party as a whole (King, Keohane & Verba 1994).

2See Gherghina & Jiglau (2011), Bochsler (2011), and Cebotari & Vink (2013), for deterministic approaches to ethnic politics.

66 In Bosnia and Herzegovina, I use the electoral precinct as the unit of analysis, resulting in a data set of over 6,000 observations. In Latvia, I use the candidate-year as the unit of analysis, which produces more than 8,000 observations. In both cases, data was collected and managed using automated web scraping technology for analysis using linear and logistic regression. In this analysis, many important factors are admittedly relegated to the error term, but the large sample sizes used in this study facilitate the precise estimation of effects for hypothesis testing. While the argument has important implications for cross-national comparisons, this study focuses on within-country variation. In this way, the dissertation differs from many prominent observational studies on ethnic voting and electoral institutions.3 Cross-country analyses are important for studies which see institutions as important explanatory variables. Since electoral institutions are country-level phenomenon, variation exists only at the country-level. However, this dissertation argues that institutions and demographics affect ethnic groups in different ways even within the same country. In order to test this argument, I rely on two separate case studies to examine the implications of my theory of ethnic voting. The theory predicts that ethnic minority voters and ethnic majority voters will respond to ethnic appeals differently because of the varying incentives each group has to demand ethnic representation. In the first set of chapters focused on the case of Bosnia and Herzegovina, I demonstrate how these incentives influence voters. Through a study of split-ticket voting in a single election, I show that voters are willing to support ethnic and non-ethnic parties even on the same ballot in ways that comport with the predictions of the theory. In the case of Latvia, I show that these effects are strong enough that they shape the over-time trajectory of the development of the party system. The first study examines voter-level implications of the theory; the second looks at the elite-level implications. Both compare the majority- and minority-group actors within a single country, rather than examine cross-national variation in political outcomes. This within-country approach is helpful because it holds many country-level factors constant. This historical relationship between groups, for example, may vary greatly between countries, affecting the types of political demands voters are likely to make of their political leaders. The two cases mentioned above were chosen for important theoretical and practical rea-

3See Rae (1967), Ordeshook & Shvetsova (1994), Neto & Cox (1997), Sisk & Reynolds (1998), Montalvo & Reynal-Querol (2005), Clark & Golder (2006), Huber (2012), and Li (2018).

67 sons. First, they share some important similarities. They are both ethnically diverse democracies. They are also both highly proportional, parliamentary electoral systems with high electoral thresholds. Ethnicity is politicized in both countries, with ethnic identities highly salient in elections and in day-to-day life. Both countries also have extremely fluid party systems, and both recently democratized after a period under a communist authoritarian regime which actively suppressed ethnic political organization. Both countries had their first democratic elections in the 1990’s as communism collapsed, and both had to create multi-party democracy mostly from scratch. Both have seen a relatively high degree of electoral volatility in the post-communist period, with new parties forming frequently. As a result, there are few constraints imposed by party system “freezing.”4 Within this highly fluid and volatile party system, voters are given viable options of both ethnic and non-ethnic parties. Reflecting both countries’ ethnic divisions and their respective communist pasts, would-be leaders attempt to mobilize voters in both countries on ethnic and non-ethnic platforms, giving voters a real choice between ethnic and non-ethnic representation. The biggest differences between the two cases are in institutional complexity and legacies of violence. While both countries transitioned from communist authoritarian states to democratic regimes in the 1990’s, Bosnia was unable to do so without first going through an extremely violent civil war. This war not only took an enormous human toll with more than a hundred thousand deaths and over two million displaced people, but destroyed much of the country’s economic base and left a lasting distrust between ethnic groups. Latvia, on the other hand, saw enormous tension between ethnic groups, as well as very serious concerns of institutional discrimination and contentious debates over historical legacies of communism, but avoided large- scale ethnic violence. As a result, Latvia’s economic and political development has greatly surpassed Bosnia’s. Latvia has joined the European Union and the Eurozone, whereas Bosnia’s membership application has been stalled by political corruption and inefficiency. Latvia is ranked as “Free” by Freedom House, while Bosnia remains “Partly Free,” and Latvia’s GDP per capita is more than twice Bosnia’s. More importantly, the nature of their relative transitions created enormous differences in the

4See Lipset & Rokkan (1967) and Mair (1998).

68 two countries’ institutional environments. In the wake of the Bosnian civil war, international mediation and peace talks guided the writing of a new constitution that would govern post-war Bosnia. This constitution may very well be “the world’s most complicated system of government (Nardelli, Dzidic & Jukic 2014).” It is decentralized into an asymmetrical arrangement of ethno-federal entities, with territorial delineation and electoral quotas at various levels of government designed to create ethnically-homogeneous self-ruling units. Bosnia’s institutional complexity is a result of international negotiators trying to build consensus among three separate wartime factions. Latvia’s constitution, on the other hand is a straightforward example of an extremely proportional and centralized parliamentary system. Whereas Bosnia’s constitution was engineered, Latvia’s was resurrected. When Latvia voted declared independence from the Soviet Union, they also voted that the communist government had been an illegally imposed occupation of Russia over Latvian territory. They therefore returned to the constitution that had been in eﬀect in 1940 when the USSR took control of the country. Bosnia’s constitution provides an opportunity to hold political issues, group identities, and political parties constant while providing variation in the institutional and demographic variables crucial to the workings of the theory. This unique institutional environment facilitates a study which has important implications for how ethnicity inﬂuences voting behavior in other countries of the world. By looking at how voters split their tickets between ethnic and non- ethnic parties in a single election, we can directly observe how voters in an ethnically divided society respond to changes in social context. Likewise, Latvia’s historical circumstances provide plausible exogeneity in institutional environment. As discussed above, the potential for reverse causality is high in studies which argue that institutions have explanatory power over political behavior. Yet Latvia’s leaders chose to impose a constitution that was written before waves of migration made Latvia the ethnically diverse country it is today. The research design in the analyses that follow do not attempt to compensate for these idiosyncratic institutions, but instead employ them in ways designed to derive more generalizable conclusions. On a personal note, the choice of these two countries also maximize my own strengths and capabilities as a researcher. Over the course of my research and education, I have spent substantial periods of time in post-communist Europe—not only Bosnia and Latvia, but also North

69 Macedonia, Bulgaria, Russia, and Hungary. I am fluent in Russian, which has enabled me to develop a working competency in reading several other Slavic languages, including Bosnian, Serbian, Croatian, Macedonian, and Bulgarian. Over the course of this project, I also developed a basic level of reading competence in Latvian, a non-Slavic Baltic language separate and distinct from the others. I have conducted dozens of formal, structured interviews with political elites in these two countries, ranging from mayoral and city council candidates in smaller towns, to political party strategists, to state bureaucrats, to the chief executive’s office. I have also ben- efited from friendship, collaboration, and hundreds of informal conversations with everyday citizens, students, and local scholars from the region. While the countries and ethnic communities remain distinct from each other, there is a degree of commonality throughout the region, stemming from shared post-communist heritage, regional historical experience, and common culture. This unique experience has enabled me to gain some cultural fluency with the politics and political discourse of the region, while also maintaining objectivity and distance.

3.2.1 Data Sources

This dissertation is a mixed-methods project, incorporating qualitative and quantitative approaches into a single study. I rely on two distinct data sources. One is quantitative voting data obtained directly from the state election administrative agencies in the two countries under study. These data were acquired using automated web-scraping technology, in order to obtain results at a level of analysis (precinct-level in the case of Bosnia, and candidate-level in the case of Latvia) which makes large-= studies possible. These figures are official election returns, and constitute the raw data used to analyze trends, patterns, and systematic variation in voting behavior. The second data source are the appeals that candidates make to voters, which is used to classify parties and candidates as ethnic or non-ethnic in nature. My measures of party ethnic identification rely on my own readings of materials made available to voters by the parties themselves. Unlike voting data, this is somewhat more subjective. In coding parties and the appeals that they make to voters, I rely on the criteria derived from the definition of an ethnic party outlined in Chapter 2. The standard employed is whether or not a reasonably intelligent

70 voter could infer from the information that the party has made available that this party purports to represent a specific ethnic group. If that answer is yes, then I code the party as ethnic, and identify which group the party claims to represent. This approach may be somewhat controversial. The most serious concern is that a research design that relies on a researcher’s interpretations of the likely impact a claim will have on its audience may suffer from concerns over measurement validity and reliability. Equally well- informed and competent researchers could reasonably disagree over how a party should be coded when presented with the same data. Moreover, there are a variety of different ways that parties could communicate their positions to voters. As I show in Chapter 4 and 6, political actors in the two countries under study often communicate their ethnic affiliation using not only overt ethnocentric rhetoric, but historical allusions and visual cues, which may not always be obvious to outside observers. Despite these concerns, I argue that such an approach is ideal in this case. First of all, the theory outlined in Chapter 2 is premised on the idea that both political elites and voters have some degree of agency in their instrumental uses of ethnicity. While they cannot define their own ethnicity or readily change it (since ethnic identities are those associated with decent), they can decide whether to accentuate those identities or downplay them; to organize along ethnic lines or some other social cleavage; or to ignore their own identities completely. The ways in which actors use those identities will be inevitably linked to the nature of those identities themselves as well as the other identities present in society. As much as possible, the measurement used to classify appeals as ethnic should be guided by the political actors themselves, and how those actors see the world around them. While it may be difficult to measure directly what a party intends to convey to voters, or how citizens understand the information given to them, it is important for the research design to take into this account, rather than imposing overly strict standards divorced from the lived experience of the actors themselves. Moreover, this approach which applies interpretative methods to the words and actions of political entrepreneurs to produce qualitative measures of ethnic mobilization is simply better suited to this research project than the other alternatives. One of these possible alternative measurement strategies would be to rely on expert surveys for codings of which parties are eth-

71 nic or not. Expert surveys on political party positioning and messaging have been enormously valuable in producing important findings on political responsiveness, coalition formation, and electoral outcomes. Expert survey data also is available for many countries in the region already through extremely valuable resources such as the Chapel Hill Expert Survey (Bakker, De Vries, Edwards, Hooghe, Jolly, Marks, Polk, Rovny, Steenbergen & Vachudova 2015, Polk, Rovny, Bakker, Edwards, Hooghe, Jolly, Koedam, Kostelka, Marks, Schumacher et al. 2017), and the Ethnonationalism in Party Competition dataset (Szöcsik & Zuber 2015). As Benoit & Laver (2006, p. 3) argue, “The great virtue of an expert survey is that it sets out to summarize the judgments of the consensus of experts on the matters at issue, and moreover to do so in a systematic way.” This is indeed an advantage of expert surveys, but it is a liability for this project. In the two countries under study, extremely volatile party systems make establishing consensus difficult. Parties which contest elections but then lose, or new parties which do not have long established reputations may be difficult for experts to place exactly on the political spectrum. In the 2007 Chapel Hill Expert Survey, experts were asked to evaluate the positions of nine political parties, despite the fact that the ballot officially contained nineteen political parties. While it is true that twelve of those nineteen political parties failed to win any seats in parliament that year, those twelve parties together took nearly 11% of the total vote in that election. Three of the parties which did win seats in that election did so with less than 10% of the vote, suggesting that those “wasted” votes taken together represent a politically substantial chunk of the electorate. The theory outlined in Chapter 2 also makes predictions about the types of parties that are likely to lose elections. Excluding parties on the grounds of irrelevance discards data which could be used to falsify the theory. I argue that parties that make ethnic appeals to groups that have little to gain by ethnic representation or are likely to win elections are the least likely to attract large levels of voter support. Excluding from the data set those parties which fail to attract voter support (and are therefore unlikely to leave a lasting impression in the minds of experts) therefore makes it impossible to empirically observe an important component of the theory. There is an additional complication in these consensus-driven expert surveys for studying small countries like Latvia and Bosnia: a dearth of experts. Hooghe, Bakker, Brigevich, De Vries, Edwards, Marks, Rovny, Steenbergen & Vachudova (2010) report that in the 2002 and 2006 Chapel Hill

72 Expert Survey, Latvia only barely met the standard of having at least four survey experts for inclusion in the final data set.5 Such issues are disproportionately likely to affect smaller, less frequently studied countries like those in this dissertation. A second existing data set that could be used is the Manifesto Project Dataset from the Man- ifesto Research on Political Representation (MARPOR)(Volkens, Lehmann, Matthieß, Merz, Regel & Weßels 2017). MARPOR’s data has been enormously useful to scholars tracking party positioning and issue salience over time. MARPOR scholars apply a line-by-line coding standard to the manifestos, party platforms, and pre-campaign policy positions released by political parties themselves to measure systematically which issues parties discuss, and which policy positions they adopt. MARPOR data is especially well-suited to comparative cross-national studies, focused on variation over time and between countries. The standardized coding scheme facilitates the comparison of party competition along a variety of axes, separating positions on foreign policy, redistribution, social values, political institutions, and other ideological positions into separate categories. However, there are two reasons not to rely on the MARPOR data for this specific project. The first is coverage. Echoing concerns with expert surveys listed above, MARPOR data does not cover the whole range of options available to voters in the countries under study. In Latvia, the MARPOR data is again limited to those parties which win political office, excluding the not insubstantial number of votes that went to those parties that failed to clear the hurdles to political representation. In Bosnia, the MARPOR data is limited to 25 parties who ran between 2006-2014. As this project requires data on those parties which may be politically relevant only at the local level but not the state level, or even those that are not relevant at any level, codings are needed for more than 100 parties in Bosnia’s extremely fragmented political system. Secondly, the focus on cross-national comparison of the MARPOR data represents more of a liability than an asset for within-country analysis. The MARPOR data is specifically designed to focus on commonalities across countries to facilitate comparative studies. For example, in Germany, France, and Britain, nationalist parties that insist on a return to traditional values and make appeals to voters on the basis of patriotism and nationalism could be coded as “Tradi-

5Neighboring Estonia did not meet this qualiﬁcation in 2002, and won the dubious distinction of being the only country dropped from the data set for lack of experts.

73 tional Morality: Positive”, and “National Way of Life: Positive”. A party in one of those same countries that argues for a multicultural citizenship, civic nationalism, and representation for ethnic minority groups could be coded as “Multiculturalism General: Positive” and “Under- privileged Minority Groups”. This assumes that support for a national way of life is essentially the same in Germany as it is in France, which for comparative cross-national studies is sen- sible. Measuring the level of support for French nationalism in Germany is largely pointless, while comparing German nationalism in Germany to French nationalism in France can produce important insights. That is not the case in this study. In the theory outlined earlier, ethnic appeals are important for two reasons: first, they establish that the political party or candidate will disproportionately redistribute benefits on the basis of ethnicity as opposed to some other social cleavage—i.e., ethnicity over class, region, ideology, etc. Secondly, they identify which group will benefit, and which will not. Classifying parties as nationalist or multicultural captures variation on the first dimension, but not the other. This is important in a country divided between multiple ethnic groups. In the context of the former Yugoslavia, for example, Croat national parties, Serb national Parties, and Bosniak national parties can all be understood to share some commonality in the nature of their ethnic and nationalist appeals to voters. But it is not just the nationalist appeal that matters: it matters which group that appeal is being made to. This variation is not only important because inter-group conflict may be the defining feature of political contestation, but also because this variation is essential to confirming or rejecting the theory. As an example, consider the National Alliance in Latvia, which is regarded by many journalistic and academic sources as one of Europe’s far-right populist parties. It is a vocal anti-immigrant, anti-Islam, anti-LGBT, ethnic nationalist party (Mudde 2014, Martyn- Hemphill 2016b, Zakharov & Law 2017). The MARPOR coding, though, includes several determinations of “Multiculturalism: Positive,” the same code assigned to Harmony, a party which dominates Latvia’s ethnic minority interests. The source of the confusion is that the National Alliance technically makes ethnic appeals to other ethnic groups in its pre-election program. Specifically, they promise that “[w]e will strengthen the role of the Latvian language, especially in the media and in commerce. We will ensure the preservation of the Livonian lan-

74 guage, the Latgalian culture, and the various dialects of Latvian.” As I discuss in the historical review in Chapter 6, this is actually an exclusionary appeal based on familiar interpretations of Latvian history. Livonian is, practically speaking, a dead language, its last native speaker having died in 2013 at the age of 103 (Charter 2013). Livonians are an indigenous people native to the northwest coast of modern Latvian territory, and thus the preservation of the Livonian language is a largely academic and cultural exercise in preserving Latvia’s ancient and pre-Soviet history. Likewise, Latgalian is considered by many scholars to be a historical antecedent of the modern standardized Latvian dialect. It is mutually intelligible with Latvian, and spoken sometimes in the Latgale region—a region with a very high concentration of ethnic Russians. While the support for the Livonian and Latgalian languages may technically be an appeal to a form of multiculturalism, it almost certainly would not be perceived as a tolerant and welcoming stance to the ethnic Russians who are unlikely to speak either of those languages, and who form the majority in many areas of Latgale. Likewise, when For Human Rights in a United Latvia, a Russian interest party, proposes that more university instruction be conducted in Russian in order to facilitate Russian employment opportunities, such an appeal is also technically a form of “multiculturalism.” Yet it is likely to be perceived as a direct challenge to ethnic Latvian nationalists, who view Latvian as an endangered language that must be preserved by state authority. ForHRUL’s advocacy of the Russian language (not a general form of multilingualism, but specifically Russian-speaking universities) is clearly a targeted message that Russians will benefit from their policies if they are elected. Likewise, the National Alliance’s appeal to Lat- vians, Livonians, and Latgallians—in other words, everyone but Russians—is an ethnic appeal to non-Russian groups. The definitions used in this dissertation and the theoretical predictions being tested require a data set that measures this variation with specific attention to local context. I therefore follow Protsyk & Garaz’s (2013) appeal to employ novel codings by specialists with regional knowledge, and to treat ethnicity as a separate axis of political competition. This approach builds on the general approach of the MARPOR dataset—i.e., that manifestos and campaign materials released by parties can be used to measure party positions—but generates codings with standards reflective of the specific goals of this research project (Budge 2003). I employ a data set broader than the MARPOR data, examining all political parties registered

75 on the ballot in the countries understudy rather than only those that win office, and specifically employ the definitions of ethnicity, ethnic party, and ethnic voting outlined earlier (Zulianello 2014).

3.2.2 Mixed Methods

This is a mixed methods project, combining qualitative ethnographic and interpretive methods with quantitative analysis. However, this is not a mixed-methods project in the sense of using multiple methods to confirm a single hypothesis (Collier & Elman 2008). In these traditionally- understood methods of “triangulation” (Denzin 2009), quantitative methods such as surveys or analyses of voting data are used to confirm or reject hypotheses, and then qualitative methods such as interviews or text analyses are used to document individual-level phenomena which further confirm and illustrate the broader findings. An example of this would be to use survey data to document which types of people support which types of parties, and then to interview individual voters to confirm that they themselves understand their voting behavior in ways consistent with the mechanisms outlined in the theory. For this particular study, this approach could take the form of using survey data to confirm that ethnic minorities are less likely to support ethnic parties when they are less likely to win elections, and then confirming through interviews that voters based their voting decisions in ways predicted by the theory. In this sense, seeing qualitative evidence that voters considered both the benefits of ethnic representation and the likelihood of actually achieving it in their voting decisions would serve as a “hoop test” that increased confidence in the theorized mechanism driving behavior confirmed by the statistical analyses. If, however, voters provided some other rationale for their voting behavior, then we might infer that the theorized mechanism is incorrect, and that omitted variable bias is driving the results of the quantitative study (Mahoney 2012). This approach would also be consistent with a mixed-methods Bayesian approach which sees quantitative and qualitative analyses as independent observations which can help alter expectations on the relationship between variables of interest (Humphreys & Jacobs 2015). Instead, my approach seeks to integrate interpretivist methods into the quantitative analysis. Drawing on the insight of Wedeen (2002), this dissertation agrees that “[i]deas, beliefs, val-

76 ues, ‘preferences,’ and decisions are always embedded in a social world, which is constituted through humans’ linguistic, institutional, and practical relations with others (Wedeen 2010).” In other words, to understand when voters choose to support ethnic parties, it is important to understand how the voters themselves understand that choice, and how the entire range of choices between parties and candidates is presented to them. This is important not only for constructing a narrative explaining the general patterns observed (Bates, Greif, Levi, Rosen- thal & Weingast 1998, Laitin 2006), but should also guide the quantitative analysis as well. The goal of the analyses in the chapters that follow is to produce a quantitative analysis of vote choice that is informed by the interpretations of the choices available to voters. In this sense, the classification of parties used in the analysis is guided by a fundamentally different motivation from comparative studies like the MARPOR. Rather than consolidating data based on universal coding standards that allows for cross-national comparison, this data used here is unique to each individual country under study, employing standards that incorporate the country’s unique historical, social, and political context. This is not to limit the conclusions of the study only to the countries in question. Instead, it takes seriously particularities of local context to make general conclusions. Ethnicity and ethnic voting as defined in chapter 2 are phenomena which exist throughout the world. This research design employs a measurement strategy which seeks to fully appreciate local nuance in order to understand universal phenomena.

77 4

Ethnic Politics in Bosnia and Herzegovina

In chapter 2, I outlined a theory of how group status could be an important determinant of voting behavior in ethnically divided societies. In this and the next chapter, I focus on the observable implications of the theory in a specific case: Bosnia and Herzegovina.1 Bosnia is an important case for a its unique historical, demographic, political, and institutional circumstances. In many ways it is an extreme case of ethnic division. It has a recent history of ethnic violence, having fought a brutal ethnic civil war characterized by genocide and ethnic cleansing. The lasting wounds of this conflict are still evident today, with distrust of ethnic outsiders and ethnic polarization pervasive throughout much of Bosnian society. Today, ethnic conflict has made Bosnia a poster child for paralysis and political dysfunction brought on by polarization (Bahtić- Kunrath 2011, Less 2016), preventing even the most basic state services from being provided to citizens (Bilefsky 2013, Deets & McClelland 2018) and stifling Bosnian economic development (Goldstein & Davies 2015). Bosnia in many ways represents a “worst-case scenario,” highlighting the most serious difficulties of democratic governance in an ethnically divided society. Even in this extreme environment, Bosnia shows enormous variation in the way political actors employ ethnic labels in mobilizing voters to win elections. Throughout history, ethnicity in Bosnia has taken on varying levels of salience, ranging from virtual irrelevance to crucial importance. Having been ruled by communists for the latter half of the twentieth century, most Bosnian citizens living today grew up under a system that ostensibly claimed that ethnic divisions were bourgeois illusions employed by aristocrats to divide and oppress the people of

1Hereafter referred to as “Bosnia.”

78 Yugoslavia. Those same citizens lived through the breakdown of Yugoslav communism and a devastating civil war where large numbers of people died as a consequence of ethnic identities they were born into and powerless to change. These citizens have seen political leaders take a variety of public positions on the role of ethnicity in public life: from ethnicity as the most fundamental social distinction guiding all political activity, to ethnicity as distraction and impediment to a just society. For this reason, Bosnia is an important research opportunity to study ethnic voting. Eth- nic divisions are pronounced and highly relevant to day-to-day life. Yet ethnic voting is hardly hegemonic, and ethnic parties routinely compete with parties that reject ethnic affiliation as a way of building linkages with voters. In this chapter, I outline the various ways ethnicity has been politicized in Bosnia. I first explore the historical background of Bosnia, describing the nature of the ethnic divisions within the country and how they have evolved. I then describe the relationship between ethnic diversity in Bosnia and its political institutions, outlining how Bosnian democracy emerged largely in response to ethnic conflict. I then provide an overview of how contemporary Bosnian political actors—principally the leadership of political parties— utilize and invest in their own ethnic identities to win elections in this particular social and institutional environment. Finally, I argue that the relationship between Bosnian political parties and ethnic identities should be understood using a three-part classification system. This primarily qualitative and ethnographic chapter seeks to describe the important distinctions between the various strategies used by political elites to mobilize voters and win elections. These descriptive classifications provide the basis of the data set used in the quantitative analysis in the next chapter.

4.1 Historical Background

Bosnia today is a country divided between three main ethnic groups: Serbs, Croats, and Bosni- aks. Oﬃcial estimates put Bosniaks at a bare majority of the country: 50.11% of the population by the oﬃcial 2013 census.2 Serbs are the second largest group, at 30.78%, and Croats are the

2The 2013 census in Bosnia—the ﬁrst since the end of the civil war which undoubtedly caused massive demographic shifts—was highly politicized, and its results remain hotly contested. See Perry (2013), Armakolas &

79 smallest at 14.6%. Popular conceptions and stereotypes hold that these three groups have long been in conflict against each other. This interpretation—not uncommon even among Bosni- ans today—grossly oversimplifies the nature of the differences between the groups, relies on erroneous essentialist assumptions, and ignores enormous variation in the nature of the three identities over time. While the distinction between the three ethnicities does have its origins in religious and cultural differences between people of the Balkan Peninsula dating back hundreds of years, these identities have changed dramatically throughout history, usually the result of political entrepreneurs responding to strategic incentives generated by political and social circumstances. In fact, the three groups share a common ethnic ancestor, speak a common language, and have been incorporated into political communities that transcend ethnic barriers numerous times throughout history. All three groups are Slavic, a subset of the same larger ethno-linguistic family that includes Russians, Poles, Czechs, and others. Bosnia’s primary ethnic groups are known as South Slavs, having a common ancestor in the Slavic peoples who moved into the Balkan Peninsula during the sixth and seventh centuries C.E. (Dvornik 1962). This common root is still evident in the languages spoken by Balkan Slavic populations. Croatian, Bosnian, Serbian, and Montenegrin have been determined by many linguists to be dialects of the same language, even if politically many politicians and citizens throughout the region maintain they are linguistically distinct (Milekic 2017). This mutual intelligibility extends to other south Slavic dialects, including Slovenian, Bulgarian, and Macedonian, all of which share the same common linguistic ancestor. The languages are not intelligible with Albanian, Greek, Romanian, or Turkish—languages which emerge from distinct non-Slavic communities who arrived in the Balkans at different historical periods. Croatian and Bosnian are written in the Latin alphabet while Serbian is written in the Cyrillic, but all three are mutually intelligible when spoken.3

Maksimovic (2014), Bieber (2013), Pasic (2013) and Toe (2016). 3Despite the mutual intelligibility of the languages, language is often used as a proxy for ethnicity and culture, and linguistic differences are politically significant. As an example of both the mutual intelligibility of the languages and the political importance of the minute differences between them, consider the packages of cigarettes sold in shops in Sarajevo. Labeling laws required that health warnings be printed on every pack of cigarettes, and one of the most common warnings is “Smoking Kills.” The phrase is literally exactly the same in Bosnian, Croatian, and Serbian, but in order to avoid the appearance of ethnic favoritism, the phrase is printed three times on each box. Such redundancy is extremely common in product packaging and public signage in Bosnia today. See Is Serbo-Croatian a language? (2017).

80 Velikonja (2003) notes that the earliest recorded forms of differentiation within South Slav peoples are on accounts sorted by religious and geographic categories. In the early middle ages “Bosnia” was a territorial designation, defining a mountainous region between Rome and Constantinople. These two imperial capitals led the conversion efforts of the Bosnian Slavs, proselytizing their respective forms of Christianity: Catholicism from the Romans to the west, and Orthodoxy from the Byzantines to the east. The religious divide is still visible in the Balkans today: Croats, Slovenes and Czechs are historically Catholic; Serbs, Macedonians, and Bulgar- ians are historically Orthodox. The medieval division of the Balkan Slavic populations into groups based on their affinity to either Eastern or Western Rome also explains why contemporary Serbian and Croatian—though mutually intelligible languages—are written in different alphabets. Having famously translated the Bible and the liturgy into the Slavonic language using what would become the Cyrillic alphabet for missionary work among the South Slavs, Byzan- tine missionaries Saints Cyril and Methodius were forced to flee the Slavic lands by German clergy committed to the “trilingual theory” that only Greek, Latin, and Hebrew languages were appropriate for religious rites (Velikonja 2003, p.23). While western religious leaders insisted on Latin as the language (and alphabet) of the church, eastern Orthodox leaders had no such qualms, and Cyrillic came to dominate the areas in the Eastern half of the Balkans. Mylonas argues that this historical division is crucial to understanding how the South Slavic people came to see themselves as different despite a common ethnic ancestor and shared language: the identities are intersectional defined by a combination of writing system, religion, and territorial location (Mylonas 2003, Chapter 1). Bosniak identity emerges much later in history, again as a result of Bosnia’s relationship to larger imperial powers. By 1463, ten years after his conquest of Constantinople and the destruction of Byzantium, Mehmet the Conqueror had subdued the region and founded Sarajevo as an Ottoman provincial capital. Bosnia was then placed under the Ottoman Millet system, a pluralistic legalism that guaranteed political autonomy and protection for some communities based on their religion. This policy made the religious distinction that had long been present in the Balkans a legal one, institutionalized by Ottoman administration (Bringa 1993). While legally protecting Christianity, Ottoman administration also encouraged conversion to Islam.

81 Administrators often showed favoritism towards other adherents of their faith, and conversion to Islam could help obtain advantageous treatment. Ottoman military conscription also facilitated the conversion of many Slavs to Islam, both by taking young children from their parents and raising them as Muslims, and by allowing provincial families to exempt themselves from service by adopting a new faith. Velikonja (2003, Chapter 4) argues that for most Slavic Christians, conversion to Islam should be understood as a matter of convenience and expediency. The distinction between the three ethnic groups therefore have their most relevant origins in religious distinctions of the middle ages and early modern era. But it is important not to underestimate the degree of social commonality between the groups. The three groups share a common Slavic ancestry, and mostly speak the same language. Given the remoteness of the region, the relative lack of clerical religious administration, and the relatively frequency of contact between the three groups, many historians have argued that religious beliefs and practices were often highly ﬂuid and changeable for most of Ottoman Bosnian history. People in this period routinely absorbed traditions from all three religions into their daily lives, possibly ignorant of the theological origins of the ceremonies they practiced or the holidays they celebrated (Donia 2000).4 The identity divides in pre-modern Bosnia were in many ways cross-cutting: as the most prominent example, Bosnian Muslims had the same religion as the Ottoman Turks, but linguistically, ethnically, and culturally were closer to Croat Catholics and Serb Orthodox Christians. The identiﬁcation along religious lines persisted through Bosnia’s incorporation into Austria-Hungary, into the era of the monarchical Yugoslav kingdom, and through World War I. Mirroring the way Ottoman policy had institutionalized religious distinctions, Yugoslav communists also enshrined ethnic and religious divisions in its policies of state administration. Yu- goslavia in the immediate post-war period was dominated by the mainly Serb Partisan faction

4Niškanović (1978) documents a clear example of how traditions of religious coexistence had persisted into modern times. In his fieldwork accounts, he narrates participating in a religious celebration on August 2, when all three main Bosnian religions religion celebrate the feast of the Prophet Elijah. Despite the different significance the three faiths place on the prophet, the religious celebrations often start out separately, but come together over the course of the day as the respective parties become larger and less orderly. Even today there is a Bosnian saying, “Ilija in the morning, and Alija in the afternoon” (“Do podne Ilija, popodne Alija”), referring to how the saint would be called by his Christian name before noon, but by his Muslim name later in the day, presumably as a result of festivities celebrating the same prophet but pronouncing his name differently converging in public places. See HadžiMuhamedović (2018), and Aura (2018).

82 who had prevailed in World War II over Croat and Muslim factions allied with the Third Reich. These Partisans maintained ties to the Soviet Union and Russian communists, and were highly inﬂuenced by Stalin’s approach to “The National Question,” which allowed for ethnic and national identities to serve as the basis of allocation and administration under a multi-national but still Marxist communist empire. Building on this model, the Partisan leadership in Belgrade designed a communist Yugoslavia as a union of separately administered homelands of various national groups composing a single state under a single communist party. The criterion for status as a constituent republic was therefore the presence of a national group. Slovenia, Croatia, and Serbia were obvious choices, with strong senses of ethnic and national identity, but the Yugoslav communists went even further than that to take what some saw as the surprising choice to acknowledge Montenegro and Macedonia (Carmichael 2015). Bosnian identities, however, were more problematic. Many Marxists saw Bosnia’s only unique identity community as primarily religious in nature—the result of ignorant Serb or Croat peasants long ago duped into adopting a foreign religious belief. Slavic Muslims were therefore initially seen as the remnant of capitalist Ottoman imperialism—anathema to communist ideology—and rather than establishing a homeland for Slavic Muslims, Yugoslav communists sought to suppress public forms of Muslim political identity by banning both Muslim youth organizations, and traditional Muslim clothing and headscarves. Borders were drawn to divide Muslim populations between the Montenegrin and Serbian republics, creating a “Bosnian” republic understood to be purely administrative, not national or ethnic in character. From the very beginning this approach seemed to ignore what was obvious among Bosnian residents of all ethnic groups—that a Muslim ethnic identity was present and strong within Yugoslavia. By 1961, the Yugoslav communist party conceded this point, and adopted a policy of treating Muslims as a distinct ethnic identity, allowing for the category “Muslim in the ethnic sense”5 to be added to legal documents (Carmichael 2015, p. 106). This was upgraded to a national identity, equivalent to Croats and Serbs in 1971 and conferring further legal rights, to “Muslim in the sense of a nationality”6 (Allcock 2000, p. 335). This both acknowledged the presence and strength of an identity culturally and ethnically distinct within Yugoslav society, but also helped to reconcile the apparent contradiction that many of

5“Muslimani u etnickom smislu” 6“Musilmani u smislu narodnosti.”

83 those referred to as “Muslims” were atheist. Despite the official acknowledgment of ethnic divisions in state administration, the prevail- ing idea of Titoist communism was “bratstvo i edinstvo” (“brotherhood and unity”). Government activities were purposefully designed with a goal of uniting various Yugoslav citizens under a multiethnic ideological platform. This policy of bratstvo i edinstvo was pursued more fervently in Bosnia than any other Yugoslav republic, most likely as the result of both the frequent interaction of members of the three ethnic groups, and the pressing state need in suppressing possibly ethnically-motivated grievances. Communists made special efforts to integrate Bosnian institutions of public and social life, including schools, factories, and military units. This attempt at multiethnic socialization and mobilization was largely successful, and by the late Yugoslav period Bosnians of all ethnic groups showed extremely high levels of social tolerance and multiculturalism. Hodson, Sekulic & Massey (1994) found in a study conducted before the collapse of Yugoslavia that Bosnia and Vojvodina were both the most ethnically diverse regions of the Balkans, as well as where survey respondents expressed the greatest tolerance towards ethnic and national outsiders. This was not only empty talk, as Bosnia also had an extremely high rate of ethnic exogamy. Thompson (1992) claims that 27% of all marriages in Bosnia were ethnically or religiously mixed.7 The long-term results of these multiethnic unions and the socialist ethnic unification policies can be seen in census data. In the 1981 census, 7.9% of Bosnians listed their ethnicity as “Yugoslav,” a civic nationalist identity tied to the state, not the traditional ethnic identities of constitutive republics (Savezni zavod za statistiku 1981). The current use of the term “Bosniak” came about during the early 1990’s, when communist rule began to collapse and previous taboos on ethnic rhetoric crumbled. As Serb and Croat nationalism became more prominent among political authorities in Belgrade, concerns that the use of the term “Muslim” to deny Bosnian ethnic and national identity became a more pressing issue

7Some scholars have critiqued this ﬁnding by suggesting that mixed marriages were actually least common in the most urban and ethnically diverse areas of Bosnia. They interpret this as suggesting that where Bosnian citizens had the options of marrying within their own ethnic community because each group was large enough to be independent, they usually would. The prevalence of mixed marriages, they argue, is driven by high rates of migration caused by economic development. In other words, Croats may be relocated to predominantly Serb areas, and when that happened, they would marry a Serb because that was the only choice available. As this happened frequently, these scholars argue that it is the migration patterns that drive the exogamy, not the underlying attitudes (Botev 1994, Bringa 1991, Knežević 2001). While the critique may suggest that ethnicity was not irrelevant to Bosnian marriage decisions in the 1980’s, it also suggests that for large parts of the country, individual preferences for endogamy were not so strong as to disqualify a potential spouse.

84 within the Muslim community. The term “Bosniak” was not, strictly speaking, new. It had been used during both the Ottoman and the Austro-Hungarian periods, but had largely been a regional term, describing Slavic people of all faiths who lived in the area understood to be Bosnia. Set against a rising tide of inter-ethnic hostilities, the 1993 Congress of Muslim Intellectuals voted to adopt the term “Bosniak” as the official name of their community (Hromadžić 2013). One of the main reasons the term had been chosen was that it completely inverted the presumed relationship between descent and religion that had existed before. The term “Muslim” implied that the members of the community had started out as either Croats or Serbs and then differentiated themselves by converting to Islam, feeding into the claim advanced by Serb nationalists that the Bosniaks were Serbs who should be a part of a greater independent Serbia to justify territorial conquest. The use of what had been a regionally-understood term suggested that the defining characteristic of Bosniak identity was the region they came from, and implied that “Catholic,” “Croat,” “Serb,” or “Orthodox” were deviations from Bosniak, since all three came from the same region and the same common ethno-linguistic ancestor.8 In short, ethnic identities in Bosnia are much more fluid over time than is commonly perceived. Serbs, Croats, and Bosniaks were a single ethnic group at their time of their initial migration into the Balkan Peninsula. Despite the divergences in identity generated by religious and political differences over hundreds of years, the three still speak the same language, albeit with different alphabets and accents. Nevertheless, based on the definitions outlined in Chapter 2, “Serb,” “Croat,” and “Bosniak” are clearly ethnic identities. While they may have initially been tied closely to religion, in their current sense they speak to descent-based attributes, classifying people based on which community they were born into. The nature and degree of politicization of these ethnic identities show enormous variation, even within living memory. For most of the 20th century everyday Bosnians lived and worked alongside those outside their own ethnic communities, and political authority was described in ideological terms, not ethnic. A large number of Bosnians saw themselves primarily in non-ethnic terms, identifying more with the

8The use of a regional term was especially appealing to secular and agnostic Muslim intellectuals, who sought to emphasize their own ethnic and national distinctiveness as one equal to the other peoples of Yugoslavia, without relying on a religious identity. Nevertheless, the use of a historical and regional term unique to Bosnia created a certain tension for Muslim populations in the Sandžak region of Montenegro and Serbia, where many had historical links to Islam, but were not regionally part of Bosnia. See Dimitrovova (2001).

85 civic identity “Yugoslavs” than with a speciﬁc ethnic group. The largest ethnic group in the country was not even called by the same name less than thirty years ago and thirty years before that was not acknowledge by political authorities to exist. Clearly, human agency can have an impact on the political and social uses of ethnic identities in the Bosnian context. The choice between ethnic and non-ethnic is not abstract in Bosnian history: both choices have been made within most Bosnians’ lifetimes.

4.2 The Wartime Origins of Bosnian Democratic Institutions

As Yugoslavia crumbled in the early 1990’s, three main ethnic parties came to power in Bosnia—the Serb Democratic Party (SDS), the Croatian Democratic Union (HDZ), and the Bosniak Party of Democratic Action (SDA). Once in power, these leaders began to dismantle the formal and informal institutions which had previously governed both Bosnia and Yugoslavia as a whole. Under these parties’ ethnic leadership, the stalemate and stagnation of the 1980’s gave way to hostility and violence, culminating in the Bosnian civil of 1992-1995. The political institutions which govern the country today were negotiated as part of the peace deal to end that war, and as such the negotiation process is a crucial starting point to understand Bosnian democratic politics today. Bosnia’s current constitution is an appendix to the treaty that ended the war, and should be seen largely as a concession to practical political expediency. The institutions which govern Bosnia today were not necessarily designed to provide for the most equitable division of power between the group or to facilitate economic and political development, but rather to put an immediate end to a disastrously destructive conflict.9 The interests and relative bargaining positions of the three main combatant groups during the war are still evident in the Bosnian constitution today. Bosnia is legally defined as a state composed of three constituent peoples who are identified by name (Serbs, Croats, and Bosniaks), and the ethnic interests of these three groups are protected by three types of institutional safe- guards. The first is territorial separation. Reconciling competing wartime demands for greater ethnic autonomy and the continuation of the Bosnian state as a political entity, the constitution

9For a more thorough discussion of the civil war, see Burg & Shoup (1999), Misha (1996), O’Ballance (2016), and Shrader (2003).

86 outlines a federal system divided into multiple layers of government. Almost every sub-national unit “belongs” to a specific ethnic group, with its borders purposefully drawn to encapsulate a homogeneous population dominated by a specific group, and legal rights guaranteed to the majority ethnic population. Despite the complications created by the geographic dispersal of ethnic groups throughout Bosnia’s mountainous terrain, subnational borders are drawn in an attempt to create clear ethnic majority groups within each unit, and to ensure that as many people as possible find themselves as part of the majority group in the political unit in which they reside. The second mechanism is a system of quotas and legal protections written into the electoral system, to ensure that elections produce parliamentary bodies capable of representing all ethnic groups. Electoral districts purposefully create regions dominated by single groups, and quotas guarantee that all three groups are represented in all arenas of state-level government. The third is a series of mutual veto powers that give leaders representing each ethnic group the ability to override any state policies or actions which they see as harming their own vital interests. In this way, Bosnia seems intentionally designed as an exemplar of Lijphart’s consociational democracy, having segmental autonomy; minority veto, grand coalition governments, and proportional representation (Lijphart 1977, Belloni 2004). The Bosnian state is divided into two entities: the “Republika Srpska,”10 and “The Fed- eration of Bosnia-Herzegovina.”11 In this way a country with three ethnic groups is divided into only two entities. This counter-intuitive outcome is the result of path-dependent processes that unfolded during the war. In 1992, right before hostilities began in Bosnia, the European Community proposed a compromise wherein Bosnia would be divided into Cantons based on a model used in Switzerland. This would have created a patchwork of self-governing territorial units under a loose confederation, but was ultimately rejected by both the SDA and the SDS.

10The name “Republika Srpska” is diﬃcult to translate exactly into English, because of a grammatical double entendre of the “-ska” ending in South Slavic languages. Adjectives modifying feminine nouns (such as “Repub- lic”) end in “-ska,” in Serbian, but so do the names of many countries, including France (“Francuska”), Turkey (“Turska”), and Switzerland (“Švajcarska”). Thus “Republika Srpska” can be translated as “The Serb Republic,” indicating an ethnic designation of the political entity, or “The Republic of Srpska,” suggesting a country whose name is “Srpska.” This is diﬀerent from Serbia itself, which in the local languages is referred to as “Srbija”. The government of the Republika Srpska often refers to itself as “The Republic of Srpska,” in English-language materials, but most western journalists and academics refer the entity as “the Republika Srpska”. I shall keep to the latter convention. 11Note that despite Bosnia-Herzegovina the country having a federal structure, the term “The Federation of Bosnia-Herzegovina” refers to the subnational entity.

87 The Bosniak leadership felt that Bosnia must continue to exist as a political unit with the same borders it had as a Yugoslav republic, while the Serbs insisted on the right of Serb-dominated territories to succeed and become part of a greater Serbia (Ramet 2002, p. 206-11). Burg and Shoup argue that the key to understanding the ultimate success of the negotiations which ended the Bosnian civil war is the relationship between the Croats and the Bosniaks, and the strategic sequence of negotiations pursued by international mediators (Burg & Shoup 1999, p. 292). The American delegation’s strategy was to first negotiate a cease-fire in the west of the country where fighting was primarily between Muslims and Croats, by establishing a joint Muslim-Croat Federation that would acknowledge the de facto territorial control of the two factions. This essentially ignored fighting between the two groups and the Serbs, and divided only part of the country into a Federation allotted to two groups, rather than addressing the concerns of all three groups simultaneously. The Croat-Bosniak conflict was seen as the most readily solvable, whereas Serbs were seen by some US officials as irremediably intransigent. Some thought that by ending the conflict between Bosniaks and Croats, both could turn their attention more effectively to the Serbs, increasing the military threat to Serbs and pressuring them to accept a brokered peace deal. Still others saw the Serb political leadership as so hopelessly stub- born that direct US military intervention would ultimately be required (Sciolino 1994). Thus the Federation and the Republika Srpska were created by two separate strands of negotiation tactics independent of each other. The final phase of negotiations, conducted in Dayton, Ohio in 1995, would ultimately be between the Muslim-Croat Federation, and the Republika Srpska.12 In order to preserve Bosnia as a single country these two entities were combined into a single country defined as a two-entity federation, whose official name would be “Bosnia and Herzegovina.”13 The state government

12Exact territorial boundaries of the entities is of importance in this dissertation only in so far as it establishes the link between ethnicity, territory, and political authority. Nevertheless, the complexities of drawing boundaries during the negotiations should not be underestimated. Many of the most vexing territorial issues remain unresolved more than two decades after the agreement was signed, resigned to a “miscellaneous” category of still ill-deﬁned political authority. See Bieber (2005) and Jeﬀrey (2006). 13Note that this is the full country name. It is not “The Republic of Bosnia and Herzegovina”, which would be in keeping with the naming conventions of all other Yugoslav successor states. Such a designation is politically controversial, as it would suggest a degree of sovereignty vested in the Bosnian state that was superior to that of the entities. In fact, some within the Serb and Croat factions had originally proposed that the country be called “The Union of Bosnia-Herzegovina”, suggesting that the resulting state was less than a full sovereign country. See Beriker-Atiyas & Demirel-Pegg (2000, p. 371).

88 in the capital of Sarajevo is responsible for external relations (foreign policy, immigration, cus- toms, border patrol, and treaty negotiations), but all other governmental functions are delegated to the entities. This power-sharing arrangement makes Bosnia one of the most decentralized countries in the world, with the entity-level authorities responsible for much of the day-to-day policy decisions that affect Bosnian citizens. The arrangement helps to guarantee that the most relevant political decisions are made in units that are much more ethnically homogeneous than the state as a whole. Most Bosniaks live in Bosniak-majority Cantons, most Croats live in Croat- majority Cantons, and most Serbs live in the Serb-majority Republika Srpska. The Dayton negotiations guaranteed political representation within the central government through a series of quotas designed to prevent any one group from dominating the others. At the highest level of the executive, the Bosnian Presidency is a three-person body legally mandated to contain one Bosniak, one Croat, and one Serb. The state legislature is bicameral, with the upper house—the House of Peoples—composed of 15 members: five from each of the three groups. The Dayton constitution also endows each of the ethnic communities with veto power over any act of parliament. A majority of any of ethnic group’s delegations to the House of Peoples can block any legislative act by declaring the proposal “destructive of a vital interest” of their ethnic community. This sets a very high bar for the passing of any state-level legislation. In theory, total unanimity on an issue in two ethnic delegations but dissent in a majority of one ethnic group could cause a proposal to fail in parliament, despite being supported by twelve members and opposed by only three.14 In the lower level of parliament, the House of Representatives, no formal ethnic quotas exist. Serbs can legally be elected in the Federation seats, just as Croats or Bosniaks can be elected in the Republika Srpska, and measures pass the lower house with a simple majority. Within the Federation, though, the seats are allotted by electoral districts that overlap with the cantonal borders, empowering-local level district majorities and increasing barriers to entry for very small and geographically disperse groups. Since entity borders and electoral districts overlap, the end result is that ethnically homogeneous districts usually elect coethnic representatives to parliament. This arrangement was successful in putting an end to the armed conflict in Bosnia—no small

14In other words, it is possible for a measure to be defeated despite being supported by 80% of the chamber.

89 achievement by any standard and one that has rightly been celebrated as a major victory for international peace-making. But the historical idiosyncrasies which built the institutions that would govern Bosnia in the post-war period have created numerous inefficiencies and contradictions. Bosniaks and Croats share a single entity, not because of common interest or ethnic similarity, but because of the sequence of negotiations during the war. Some commentators have argued that the legal link between ethnicity and sub-national political units enshrined by the Dayton constitution effectively legitimizes genocide by rewarding a military strategy which used war crimes as a way of solidifying political power (Bilefsky 2008). Moreover, while the legitimacy of the constitution often rests on the assumption that people have been geographically sorted into ethnically homogeneous sub-national units, the mapping between ethnicity and location is far from perfect. While most Bosnian Serbs do live in the Republika Srpska, there are substantial numbers living in the Federation—over 56,000 according to the 2013 census—where they are legally defined as outsiders. Likewise, over 200,000 non-Serbs live in the Republika Srpska. All told, the most recent census estimates that 86.7% of Bosnian citizens live in “their” entity—i.e., Serbs in the Republika Srpska or Croats and Bosniaks in the Federation. While this number is relatively high, it nevertheless shows that more than 13% of Bosnians are minorities within their entity. This is to say nothing of the Bosniaks or Croats who find themselves living in Cantons dominated by the other ethnic group within the Federation, or the ethnic minority inhabitants of highly homogeneous municipalities who live in cities dominated by another ethnic group. The entity system also completely ignores Bosnian citizens who do not belong to the three main ethnic groups, including Roma, Jews, Turks, Poles, and Albanians. Nor does it provide any legal protection for mixed-ethnicity or Yugoslav citizens, who will by definition be outside the dominant group wherever they go.

4.3 The Evolution of Bosnian Political Actors under the New

System

The three main ethnic parties—the HDZ, the SDA, and the SDS—remained important and even dominant political factions following the war. But in the decades after the peace accords were

90 Eﬀective Numer of Parties in Bosnian Lower House 9.92 10

8.58 9.42

7.12 ENP

5.86 4.44

1996 1998 2002 2006 2010 2014 Year

Figure 4.1: Bosnian Eﬀective Number of Parties 1996-2014

signed many new parties have entered the political arena, and as Bosnian democracy has evolved the political party system has fractured and diversified. Figure 4.1 shows how in the period between the Bosnian civil war and the 2014 election, the effective number of political parties in the lower house of the Bosnian parliament more than doubled.15 The dramatic rise in the effective number of parties is likely a result of the relative ease of party formation and the low barriers to entry in the various parliaments of the country. While the Dayton constitution creates overlapping layers of governance in a complicated, ethnically- segregated federalism, elections are highly proportional with extremely high district magni- tudes, relatively low electoral thresholds, and few barriers to entry for candidates. Bosnia uses open list voting (closed list until 2002) to elect the state parliament, the entity parliaments, and the cantonal assemblies. The electoral threshold for entering parliament is 3%. This makes it easy for even small groups and niche parties to obtain some representation in government. Party registration and campaigning laws in Bosnia also are highly permissive for new entrants and competitors. Political parties require registration to contest elections, for which the most

15Calculations in Figure 4.1 follow the modiﬁed Herﬁndahl index outlined by Laakso & Taagepera (1979). Data was obtained from the Bosnian Central Election Commission.

91 onerous obstacle is the collection of voters’ signatures. Three thousand signatures are required for contesting the state-level parliament, 2,000 are required for contesting the entity-level elections, and only 500 are required for contesting a cantonal election. This puts the diﬃculty of forming a new political party in Bosnia roughly in the middle by the standards of other European parliamentary democracies (LegislationLine 2006).16 The increase over time is most frequently the result of intra-party splits and leadership disputes within existing parties. Such disputes are extremely common in post-war Bosnia. Despite shared support of extreme tactics during the war Serb leaders Radovan Karadžić and Biljana Plavšić were unable to agree on post-war policy for the Republika Srpska, leading to Plavšić’s departure from the SDS and founding of the Serb National Alliance (SNS) in 1997. The Croa- tian leadership notoriously split in 2006 over whether the HDZ should adopt a moderate stance in order to appease the international community (including the increasingly moderate government of Croatia), or risk international isolation by increasing its demands for greater autonomy for ethnic Croats. Božo Ljubić led a walk-out of HDZ representatives and founded the “HDZ- 1990,” named because it sought to return to the more extreme policies the HDZ had advocated in the lead-up to the war. Within the Bosniak community, several prominent Bosniak leaders left the SDA following the war to form their own political parties. Bosniak academic Haris Silajdžić had originally been a member of the SDA and a part of the Bosniak delegation to Dayton during the peace accords, but left the party after the 1996 elections to found the SzBiH (“Stranka za Bosnu i Hercegovinu”, Party for Bosnia and Herzegovina) which has taken a harsher stance against the ethnic division of the country enshrined in the constitution than the SDA. Silajdžić has advocated for the abolition of the Republika Srpska, and denounced the SDA as too willing to absolve Serbs of war crimes in the name of inter-ethnic negotiations. The Bosnian party system is so permissive that it is not uncommon to have splits-within- splits. Fikret Abdić, the Bosniak head of Yugoslav Bosnia’s largest agriculture companies,

16Voter signatures are required in many European parliamentary systems, with substantial variation in the numbers required. Romania, another ethnically divided post-communist parliamentary democracy requires 25,000 signatures for a party to be placed on the ballot, which is roughly the equivalent of 5,000 voters in Bosnia, after adjusting for diﬀerences in the country’s population. Denmark, another highly proportional parliamentary elec- 1 toral system—albeit one with a much longer-established democracy— requires 175 of the voters to sign a petition in order to register a political party. Applied to Bosnia, this standard would increase the signatures required from 3,000 to over 10,000. See Pilet & van Haute (2012).

92 famously split with the Izetbegović government during the war to ally with ethnic Serbs (Christia 2008). His DNZ party (“Demokratska narodna zajednica”, Democratic People’s Union) met with some success in the post-war elections in the northeast of the country where his company was based. But following a series of intra-party disputes, many of the party members followed his daughter, Elvira Abdić-Jelenović, into the rival Labor Party of Bosnia and Herze- govina, which has since surpassed the DNZ in its electoral success. The tendency to resolve intra-party disputes by forming new parties is also seen in non-ethnic parties. The Democratic Front (DF), one of the most important explicitly non-ethnically identified parties in Bosnia today, was founded after a split in the communist successor Social Democratic Party. Željko Komšić, an ethnic Croat, had built up enough personal electoral support after being elected to the Croat seat of the Bosnian presidency running as an SDP member that he was able to form his own party to contest elections in 2014 and 2018. Overall, the dominance of the three main ethnic parties and the primary communist successor party has persisted throughout the post-war period. But this dominance is not absolute, and a large number of new parties have entered into Bosnian politics at all levels of government, leading to an extremely fractured political party system. The result is that political actors from the first Yugoslav election and the civil war are still highly relevant and influential in Bosnian politics, but overall voters have many more parties to choose from when voting. Some of these new parties are explicitly ethnic, and compete with other ethnic parties for the votes of their own ethnic communities. Other parties are not, and seek to build multi-ethnic bases of support in order to win elections. How exactly political parties in Bosnia employ their ethnic identities in winning elections is a subject I turn to in the section.

4.4 The Political Use of Ethnicity in Bosnian Elections

In Bosnia, many parties choose to use ethnic identities to convince voters to support them. They present their party as champions of ethnic interests, and invest substantial resources in money, time, and energy convincing voters of their own ethnic credentials. Other parties emphasize other, non-ethnic characteristics of their leadership and policy proposals. Some explicitly de-

93 nounce parties that rely on ethnic labels to attract support. These parties put up websites in multiple languages and recruit candidates from a variety of backgrounds to demonstrate to voters that citizens of all ethnic groups stand to benefit from their leadership. To understand when Bosnian voters choose parties that rely on ethnic labels and when they support parties that do not, it is important to understand and catalog the variation in the ways parties employ ethnicity to attract voter support. In this section, I catalog the various methods by which Bosnian political parties employ ethnic labels during campaigns. On the assumption that voters are “cognitive misers” (Fiske & Taylor 2013[1984]), I start from the most readily available and least costly form of information available and proceed to the most demanding. First I examine the names that political parties in Bosnia give themselves, then the graphic representations they use to portray themselves in campaign materials, and finally the texts and policies they propose in their efforts to win elections. For each, I catalog the ways that ethnic identities are employed, rejected, or ignored by various political actors.

4.4.1 Party Names

One of the most straightforward ways a party communicates to voters is its name. As it is essentially impossible to talk or think about an organization without using its name, the choice of party name is an important signal to voters about what the party stands for and how it wants to be perceived. A party name is also one of the only pieces of information that is communicated to voters by the process of voting itself. A voter who knows literally nothing about the politics of her country or who had never read anything during the campaign would still be presented with a list of party names when given her ballot on election day. The choice of party name is not only a crucial one for candidates, but given the high frequency of political party splits in Bosnia, it is a choice many politicians make more than once in their careers. Several parties mention the name of a speciﬁc ethnic group in the name of the party. The most prominent example of this are the Serb Democratic Party, and the Croatian Democratic Union, but several smaller parties fall into this category as well. These would include the Bosniak People’s Party, a minor contender with ties to Bosniak populations living in Serbia and Montenegro which never managed to challenge the SDA’s dominance of the Bosniak vote.

94 On the Croat side, many parties identify their Croatian identity by including the word “Croatian” in their name. In addition to the HDZ, this would include the Croatian Party of Rights and the Croats’ Union. Croat politicians also have a tradition of choosing party names which not only include the name of the ethnic group, but recall historically important political organizations in Croat and Croatian history. The above-mentioned Croatian Party of Rights takes its name from the “Party of Rights” formed in 1861 to advocate national unity and autonomy for Croats within the Hapsburg empire. Other parties choose names which specifically declare a non-ethnic identity. Many of these parties take names emulating the ideological parties common in Western Europe. This would include the Liberal Democrats, the Socialists, the Bosnian Greens, the Communists, the La- bor Party and the E-5 European Ecological Party. While other parties may make their ethnic affiliation clear in their names, these parties make identifying specific ethnic allegiance more difficult. Mirroring this strategy of avoiding ethnic labels, some parties which are competitive only in specific parts of the country have adopted names based on regions. The Party for a European Sarajevo competes only in the capital city, and the Posavina Party contests elections only in the northeast of the country along the Sava River. A very small set of parties have chosen names that align with individual-level identities. These parties speak to concerns of voters which are not linked to region, ideology, or even class, such as the Party of the Handicapped, the Bosnian Pensioner’s and Retiree’s Party, and the Women of Bosnia. While these parties rely on individual identities, these identities are not strictly ethnic, since they speak to individual- characteristics that are the result of circumstance, choice, or chance. Still a third set of parties has chosen names that convey very little information, opting for a kind of inoffensive banality. Several large and political relevant parties fall into this category, including the Democratic Front, the Democratic People’s Alliance, Our Party, the Party of Rights and Development, Union for a Better Future, and the People’s Party for Work and Bet- terment. Most voters would be unable to identify the priority of these parties based on names alone beyond their general support of universally popular ideas. Despite being the most dominant Bosniak party and one of the most successful parties in Bosnia, the SDA could fall into this category, as could the Alliance of Independent Democrats, the party that would come to

95 dominate the ethnic Serb vote in Bosnia in the first decades of the new century.17 Another frequent characteristic of Bosnian political parties is the inclusion of an individual’s name. On the Bosniak side, this is the case with Sefer Halilović, whose political party is officially named “The Bosnian-Herzegovinian Patriotic Party - Sefer Halilović.” Halilović was the head of the Army of the Republic of Bosnia and Herzegovina during the war. His party formed immediately after the war to contest the first post-war Bosnian election in 1996. Form- ing an independent party may have been a reflection of Halilović’s long-standing disagreements with Alija Izetbegović and the SDA leadership, as their disputes in 1996 were widely known. The name itself refers to the “Patriotic League,” an ethnic Bosniak paramilitary group separate from the Yugoslav armed forces that Halilović had founded in the lead-up to the civil war, and which would eventually evolve into the Bosnian Army. Due to this militant defense of ethnic Bosniak interests during the tumultuous time leading up to the war, Halilović is widely known and respected among segments of the Bosniak population. Among Serbs and Croats, he is most often associated with a war crimes indictment in The Hague in connection with a massacre of Croat civilians and prisoners of war in the villages of Grabovica and Uzdol in 1993. Halilović was ultimately acquitted of the charges, on the grounds that while the massacres were in fact war crimes, Halilović personally did have command authority over the troops who committed them and was not legally culpable. Halilović’s supporters have praised him as an ethnic champion who defended Bosniaks against outside aggressors, but also on the grounds that his voluntary surrender to The Hague and cooperation with their inquiry demonstrated a commitment to international norms and human rights. His supporters—almost exclusively ethnic Bosniaks— contrast Halilović’s cooperation with the war crimes tribunal and ultimate acquittal to the many other indicted military leaders who fled, went into hiding, or were ultimately convicted of numerous atrocities.18 The naming of the party seems designed to build a party identity around the personal brand of its leader. Calling a party “Patriotic” conveys little information, but attaching

17The SDA actually had its founding convention at a time when explicit ethnic political campaigning was illegal in Yugoslavia. This provision was eventually declared unconstitutional by the Constitutional Court, having been in eﬀect for only a few months. The timing of the SDA founding convention is the reason that it is the only party of the three main wartime combatants not explicitly named after the ethnic group it represents. See Burg & Shoup (1999, p. 46-8). The Alliance of Independent Social Democrats (SNSD) started as a non-ethnic party in the Republika Srpska, but has evolved over time to become the leading voice of Serb nationalism and separatism in Bosnia today. See Section 6 of the next chapter. 18See Selimbegović (2001), van der Wilt (2007), and Ramet (2012).

96 Halilović to it conveys a certain commitment to both the integrity of the Bosnian state, and the defense of ethnic Bosniaks. Another instance of personal branding being used to develop a party identity is the previously-mentioned SBB. Technically, the party registers under the name “Union for a Better Future - Fahrudin Radončić,” so that the name of the party founder shows up on the ballot and all government records. Radončić is a media mogul, founder and owner of the Dnevni Avaz newspaper. Sometimes referred to as the “Bosnian Berlusconi19,” Radončić is famous for his personal wealth and outlandish personality. The Dnevni Avaz was initially formed with support from the SDA and the Bosnian military during the war, and the paper has long been seen as promoting an ethnic Bosniak agenda. Nevertheless, Radončić seems to have broken with the SDA in recent years, forming his own party and increasing negative coverage of the Bosniak political establishment in the media outlets he controls. His decision to attach his own name to the party he founded most likely represents an attempt to parlay his own personal brand as a successful business man who has amassed a large fortune to that of his party. Some of the parties which explicitly identify as Croat discussed above also employ this branding technique of attaching the name of an individual to the name of their party, but rather than the name of their party founder they use the name of a historical figure associated with a specific ideology. The Croatian Peasant Party of Stjepan Radić invokes the name of a famous Croat nationalist who is seen as organizing the nineteenth century agrarian ethnic Croat population into a political force. Radić opposed Yugoslavism and is considered by many Croats as a martyr, having been assassinated on the floor of the Legislative Assembly of the Kingdom of Serbs, Croats and Slovenes (the predecessor to the Yugoslav state) in 1928. Radić has been a polarizing figure throughout the twentieth century, but is employed in the party branding to convey a specific and relatively extreme approach to the defense of Croat ethnic interests. Another Croatian break-away party, the Croatian Party of Rights Dr. Ante Starčević adopted the name of a famous Croatian intellectual who died nearly a full century before the party was founded in 2010. Starčević had founded a similarly-named party in 1861, which argued for Croatian independence from the Hapsburgs. The party has aligned itself in coalitions with other Croat ethnic

19See Mujanović (2014).

97 parties, but often presents itself as more extreme in the defense of Croat national interests, and is often considered by international observers to be a far-right party. In both of these instances, the choice of naming the party after a revered figure from history signals a kind of purity. Radić and Starčević are important historical figures, so much so that they both appear on banknotes in Croatia. Employing the names of these important historical figures in the party name is a way to signal that the party honors those who struggled in the past for the interests of Croats, and presumably shares those values.

4.4.2 Party Logos

Graphic logos of a political party can be displayed at campaign events as a way of building brands, and linking parties to policies and outcomes in the minds of voters. Since observing a logo requires virtually no effort, voters are much more likely to engage with a visual logo than they are to read a party platform, or listen to a campaign speech. For a disinterested voter, a logo on a public billboard may be the deepest communication that voter has with the party itself. Logos can be especially important in a new democracy like Bosnia where parties and candidates are unfamiliar to voters, and underdeveloped media makes acquiring information more costly. Linking political affiliation to visual cues is common not just in Bosnia, but throughout the democratic world: witness the United States’ perpetual discussion of “red states” and “blue states,”20 or the colorful language used to describe Germany’s “red-green” or “traffic- light” coalitions.21 In India, where illiteracy has been a major complicating factor in elections since independence, political parties are assigned mandatory symbols representing them on the ballot.22 Many Bosnian political parties have adopted logos and emblems that incorporate historically or culturally relevant images in order to convey ethnic positions immediately to voters. Figure 4.2 shows a sample of graphic logos for each of the three main ethnic parties. Figure 4.2a is the

20Red states are those that support Republicans; Blue states are those that support Democrats. The color scheme employed in the United States stems from the norms of election night coverage by the major American networks. Confusingly, it is the opposite of the color scheme used in most of the rest of the western world, where leftist parties adopt the color red, and conservatives adopt the color blue. 21Red is the color of the Social Democratic Party logo in Germany, while the Free Democrats use the color yellow to represent their liberal ideology. When these two parties enter into a coalition with the environmentalist Green Party, the three colors of their logos resemble a traﬃc light. 22See Iwanek (2016).

98 a b c

Figure 4.2: Bosnian Ethnic Party Logos logo of the SDS. The red, blue, and white crest is a reference to the Serbian flag, and the seal in the middle is a stylized version of the Cyrillic abbreviation for the name of the party, “СДС”. The text is telling, as the Cyrillic alphabet is the main distinguishing characteristic between Serbian and the mutually intelligible Bosnian and Croatian languages. Figure 4.2b shows the logo of the HDZ. In a subtle touch, the small red square above the “H” is a nod to the red-and-white checkerboard pattern of the šahovnica, the Croatian coat of arms and national symbol. In the context of Bosnia, where ethnic divisions have been reinforced by violent conflict and citizens are extraordinarily sensitive to declarations of ethnic identification, these details are impossible to miss. The use of overt ethnic imagery and iconography, in conjunction with the use of the ethnic group’s name in the very name of the party quickly eliminates any doubt which ethnic constituency this party is seeking out, and makes the ethnicity central to the party’s presentation of itself to voters. Figure 4.2c shows the logo of the SDA, which as mentioned previously does not explicitly refer to any ethnic group in its party name. Yet any ambiguity as to which group’s interests the SDA champions is clearly resolved by the SDA logo, which appears prominently at political rallies, and is used in printed and online campaign materials. The party’s fleur-de-lys blazon is a medieval symbol, dating to the 14th century pre-Ottoman rulers of the Kingdom of Bosnia. This symbol was not used during the Ottoman period, when Bosnia was often represented by explicitly Islamic symbols, nor was it common during the Yugoslav period, when a uniform

99 a b c

Figure 4.3: Bosnian Non-Ethnic Party Logos socialist iconography was imposed on all the republics. Bosniak troops resurrected the medieval herald as combat insignia during the civil war, as it was the banner of the previous rulers of an independent and sovereign Bosnia. During the war, the fleur-de-lys pattern became synonymous with ethnic Bosniak interest, and a highly polarizing symbol. It is so contentions that keeping the fleur-de-lys off the official flag of the newly formed country became an important point of contention during the Dayton peace discussions for the Croat and Serb delegations (Pauker 2012). To further reinforce their connection to the Bosniak community, and the long-standing connection between Islam and Bosniak identity, the stylized shadow surrounding the fleur-de- lys is a reference to an Islamic crescent moon, seen on the flag of many predominantly Muslim countries like Algeria, Turkey, and Pakistan. The official logo of the SDA features two green bands running along the top and bottom, a color commonly used throughout the Muslim world as a symbol of the Islamic faith. Logos are not only used to convey ethnic affiliation; they are also used to reinforce identities which are explicitly not ethnic in nature. Parties can choose logos that rely on the Yugoslav communist past or symbols common in other European countries to avoid identifying the party with a single ethnic group and project an image based on ideology or multiethnic identity. The logo for the SDP is shown in Figure 4.3a. The image recalls the red rose that has long been the symbol of the Socialist International and is still used by center-left parties in France, Italy, and other countries. The use of the color red, inherited from the League of Communists of Yu-

100 goslavia and common among labor movements worldwide, not only avoids identification with a specific ethnic group, but links the party to its ideological past and with labor movements and leftist parties throughout Europe. The logo of the Democratic Front seen in Figure 4.3b features a shield of blue and orange. Orange has no strong historical association in Bosnia, and its use is likely a reference to the Orange Revolution in Ukraine in 2004-2005. DF candidates have referred to their party as “a revolution in orange,” and DF campaign rallies have featured orange banners and flags in a way that vaguely suggests the historic protests in Kiev. The blue field with white stars is a reference to the Bosnian state flag. While the use of a flag by a politician might seem completely natural to western observes, it is important to remember that the current flag of the Bosnian state was imposed on the country by the international community less than 20 years ago. It is a symbol of the legitimacy of the multiethnic state of Bosnia and Herzegovina as it currently exists–hardly an empty gesture in a country where many still argue in support of secession. Moreover, the DF official logo has two versions: one in the Latin alphabet, and one in Cyrillic, suggesting the party’s desire to appeal to both the Latin-alphabet using Croatian and Bosnian speakers in the Federation, and the Cyrillic-alphabet Serbian speakers in the Republika Srpska. Figure 4.3c shows the logo of the Party for Bosnia and Herzegovina. The use of blue, yellow, and white is yet another reference to the county’s current flag, and the triangular design refers to the geographical shape of Bosnia’s current borders. By making the border of the country the first thing voters see when identifying the party, its leaders quickly convey that the party supports the multi-ethnic state. The letters in the logo are not an abbreviation of the party’s name, but rather the Serbo-Croatian preposition meaning “for.” This logo is highly effective in quickly communicating the central political platform of this party—support for the Bosnian state as it exists currently. If voters are cognitive misers, then political parties need to be able to quickly and easily communicate with voters. Many Bosnian parties use their logos as a part of a communication strategy. Ethnic parties often incorporate ethnic and religious symbols which have long-standing ties to specific groups in order to broadcast an ethnic identity to potential supporters. In a country where ethnic identities are highly salient and voters are sensitive to markers of ethnic affiliation, these graphic depictions serve to avoid confusion about who will likely benefit from the party’s

101 policies, and uses polarizing imagery to instantaneously differentiate in-group members from out-group members. Other parties use more recent symbols to distance themselves from ethnic identification to convey a non-ethnic identity. Images tied to a Western European identity or a communist identity avoid declaring support for a specific ethnic constituency, and references to the internationally-brokered Dayton Peace Accords suggest a civic identity based on Bosnian statehood rather than ethnic affiliation. In all of these cases, political elites are using the graphic representations of their parties to quickly tell their potential voters where they stand on these issues.

4.4.3 Campaign Rhetoric

Of course, voters don’t base their voting decision exclusively on names and logos. Office-seekers and parties interact much more directly with citizens, as well. They give speeches and media interviews, they release statements, and they make proposals. Those who win build reputations by enacting policies and administering state offices. Politicians may not have the ability to force voters to pay attention to their campaigns, but they can decide what to put out for those voters who do choose to seek out information. In Bosnia, party platforms and campaign rhetoric often follow the same patterns observed in party naming and logos. Some parties invest in explicit ethnic identification, overtly portraying themselves as ethnic champions focused on representing a single ethnic group at the expense of others. Others not only avoid explicit ethnic identification, but also attack the mobilization of co-ethnics as a harmful practice and a reason to support one party instead of its ethnic competitors. The degree to which parties employ ethnic identification in building connections to potential supporters varies widely in Bosnia. The Bosniak SDA has perhaps the most developed official platform of the three largest ethnic parties in Bosnia. This platform sets policy priorities and goals that are frequently taken as an opportunity to broadcast ethnic allegiance. In foreign policy, for example, the SDA has historically set EU ascension and integration into international organizations as a top priority, but for reasons that are particular to its ethnic constituency and not the country as a whole: “Our foreign policy priorities are full membership in NATO and the EU... In all international and domestic institutions, special attention will be paid to fighting the anti-Bosniak lobbying taking

102 place in the world today” (Stranka Demokraske Akcije 2014, p. 4). One might think that regardless of ethnicity, all Bosnians—as citizens of a very small country—might stand to benefit from collective defense, freedom of travel, and access to EU markets. Yet the SDA emphasizes that these benefits are means to advance the ethnic interests of the Bosniak community. Likewise, the section on economic policy links development and refugee return—issues which the platform claims cannot be resolved until the issue of jobs for Bosniaks is solved first: “Sustainable refugee return cannot be achieved without sustainable economic development, for which we must urgently address the problems in public education and government employment in those areas where Bosniaks are in the minority” (Stranka Demokraske Akcije 2014, p. 7). In these examples, even the most public of goods (i.e., development and security) are seen as benefits that the party will direct only to Bosniaks at the expense of others. As the platform is worded now it could even be construed as a direct attack on the other two ethnic groups, particularly Serbs. “Areas where Bosniaks are in the minority” can be nothing else but a reference to the Republika Srpska and the Federation’s Croat Cantons, just as the “international anti-Bosniak lobbying” is likely to be interpreted by many Bosnians as an attack on Serbian support for an independent Serb entity. Thus even issues with the potential to cross ethnic boundaries in their popularity are used to identify with a specific ethnic group. The HDZ campaign rhetoric also makes its investment in Croatian ethnic identity overt, obvious, and clear. In the section on defense policy, the party writes:

The HDZ believes that the defense and security policies of the state should include integration into international security processes and institutions such as the Part- nership for Peace, the Euro-Atlantic Council, and NATO.

The armed forces of Bosnia and Herzegovina must reﬂect and maintain the representation of constituent peoples and permanently ensure the recognition of the Croatian Defense Council as one of the components in rationally sized armed forces in accordance with economic possibilities and realistic security assessments....

The HDZ considers the preservation of all vital of the national interests of the Croa- tian people as an integral part of the defense and security strategy at the level of Bosnia and Herzegovina. (HDZ 2007, p.9)

103 This suggests that in foreign policy, the HDZ is in agreement with the SDA: Bosnia’s defense interests are best served by integration with European institutions of collective security. But the writers of the HDZ platform go out of their way to show that this policy decision should be made in a way that reflects Croat interests. Despite agreement with Bosniak politicians, the HDZ specifically mentions the Croat Defense Council, the main Croat military force during the war, as one that has a role to play in contemporary Bosnian defense policy because of its importance to protecting ethnic Croats in Bosnia. The HDZ will represent only the Croats in Bosnia, which they make clear throughout their policy proposal. Ethnic Croat parties often employ a strategy of attacking the Dayton constitution on the grounds that is unfair to Croats specifically. One frequently recurring grievance is that there are three constituent peoples in Bosnia but only two entities. Croats must share their entity with Bosniaks, and are thus perpetually disenfranchised because the Bosniaks are a larger community. Many Croat parties, explicitly attack the current consociational arrangement as unfair, and attribute the relative dysfunction and paralysis of the Bosnian state to the fact that not all groups are equally represented. Dragan Čović, President of the HDZ, has repeatedly claimed in television interviews that the current two-entity system is unfair and an impediment to economic development and European Union accession. In a 2017 interview with the Croatian- language television channel Hrvatska radiotelevizija, Čović argued not only that the Dayton Accords themselves may have been fundamentally unfair to Croats by not providing them with an entity of their own, but that the future of Bosnia lies in the devolution of the country into three, or possibly even four, five, or six separate entities (Oslobođenje 2017). Advocating for a constitutional change theoretically could be a non-ethnic issue, advocated by ethnic and non-ethnic parties alike. But in keeping with the logic of ethnic party, these policy proposals are always espoused because of the impact that they will have on Croats. The official party statute of the Croat Party of Rights, for example, states:

1. We are ﬁrmly committed to a European Bosnia and Herzegovina, in which the Croatian people—together with the other two constituent peoples—will enjoy full sovereignty and freedom, and where all citizens in the entire territory of Bosnia and Herzegovina will be equal. We believe that this commitment can

104 only be achieved through a complete constitutional reform and the adoption of a new Bosnian Constitution where governing institutions are under the constant control of the public.

2. We maintain that the ﬁnal solution of the Croatian national question in Bosnia and Herzegovina and the possible reorganization of the Federation should be achieved through constitutional change. (HSP BiH 2015)

This statement is a fairly specific policy proposal stance by Bosnian standards: it proposes to reorganize the Federation through the process of constitutional amendment. But the reason to amend the constitution is not because it will make the government more fair or efficient for everyone, but because it will make it fairer for Croats. Concerns of other groups, or even of various groups within the Croat community, are ignored completely. In this way, the Croat parties use language and rhetoric that make it very clear to their voters that they consider coethnics more important than ethnic outsiders, and will govern the state accordingly. Ethnic parties like the HDZ and the SDA actively emphasize their ethnic identities by viewing all policies through the lens of how it will affect their coethnic communities. Even in areas where the two parties agree on policy, they use divisive language which separates insiders and outsiders. Other parties, however, actively seek to distance themselves from ethnic identification. These parties use language which not only attack their ethnic party competitors, but also the fundamental logic of ethnic representation itself. The most prominent example is the SDP. Their party program identifies itself not as the party of any single ethnic group, but as the rightful heir of Tito’s multiethnic and internationalist socialism. The SDP declares that Bosnia is inherently a state of multiple ethnic groups, and that the post-war constitution must be preserved and upheld. Its official position on nationalism and ethnic interests is somewhat reminiscent of twentieth- century communist dogmas: “Nationalism, as well as neoliberalism and various forms of fun- damentalism, are the biggest enemies of state development, social democracy and the social democratic vision of politics in Bosnia and Herzegovina” (Socijaldemokratska Partija BiH 2014, pp.29–30). The party’s vision blends western European Social Democratic Movements with Yugoslavia’s unique twentieth-century experience of inter-ethnic cooperation. In terms recall- ing communist propaganda, party members are committed to persevere “in the struggle against

105 Figure 4.4: Bosnian Campaign Advertisements ethno-nationalist concepts of governance as well as social injustice which threatens the disintegration of Bosnian society and the Bosnian state” (Socijaldemokratska Partija BiH 2014, p.31). In this way the party links its current struggle for economic development, European integration, and effective governance to the struggle of Tito against fascism and Stalinist totalitarianism (Socijaldemokratska Partija BiH 2014, p.7). This hostility towards ethnic mobilization extends into the party’s electoral campaigns. Fig- ure 4.4 shows a comparison of political advertising done by the HDZ, SDA, and the SDP. The HDZ ads and SDA ads are remarkably similar: both feature portraits of their party chairman. Both use divisive ethnic symbols. In the HDZ ad, Čović is standing in front of a red and white checkerboard to represent Croatian national identity, whereas Izetbegović’s ad uses the color green and both the Bosniak fleur-de-lys and the Islamic crescent moon in the party logo. The language used in the ads is highly polarizing: the HDZ ad text translates to “Nation and Home- land”, clearly indicating that the policy priorities of the HDZ are understood in ethno-national terms. The SDA text reads “The people know,” suggesting a double meaning. The text implies not only that the (Bosniak) people are the ones who know how to run the country, but also that Bosniaks know that the only real vote choice is Izetbegović and the SDA. The SDP ad, however, is markedly different. The SDP has chosen not to put a single individual on their campaign poster, but a committee. They have also written all of the portrayed candidates’ names on the list, for a very clear reason: the names convey ethnic markers which suggest a multi-ethnic

106 group of people.23 The campaign slogan translates as, “A country for all of us!” suggesting that the party is refusing to align itself with a single ethnic group. The SDP is not the only party to denounce ethnic mobilization. The more recently formed Democratic Front also uses anti-ethnic rhetoric throughout its campaign materials and policy documents. While the DF has been less directly socialist, it still uses the language of universal populism to strike a markedly anti-ethnic stance. The party manifesto explicitly denounces “[t]he ethno-predatory capitalism imposed on the social life of Bosnia and Herzegovina after the war...” as “... based on exploitation, marginalization, and discrimination.(Demokratska Fronta 2014)” In domestic matters, the DF advocates for political restructuring of the state guided by European principles of efficiency and economy, rather than ethnic self-determination. The DF manifesto also outlines a multi-ethnic vision by making explicit references to all groups as important parts of Bosnian society. Its vision of pluralism is founded on principles of secu- larism and separation between church and state, and outlines a vision of recent Bosnian history that “[r]espect[s] everyone’s sacrifice and everyone’s loss...” while holding a responsibility “to find the strength to look to the future, not the past.” In this way, the DF acknowledges the impact of the civil war on contemporary politics while trying to avoid alienating any particular group. In foreign policy, the DF advocates integration into western European institutions like NATO and the EU, noting that other countries in the region have already or soon will join NATO, mentioning by name Croatia, Albania, and Montenegro. The mention of these three countries can hardly be accidental, but used to further justify the multi-ethnic appeal of NATO in the region: Croatia, obviously, is the “motherland” of Bosnia’s ethnic Croats; Albania is a historically Mus- lim country with an Ottoman heritage shared by Bosniaks; and Montenegro is a predominantly Eastern Orthodox country with close cultural and linguistic ties to Serbia. Some parties employ similar anti-ethnic rhetoric in their programs, but emphasize regional and local issues, rather than international. The Posavina Party employs language of “polycen- trism” and “anti-fascism” to explicitly denounces the ethnic organization of the state, claiming that the Dayton constitution has created “permanent crisis” that benefits only the kleptocratic

23“Čengić” is a Slavicization of a Turkish name linguistically related to “Genghis” that would be interpreted by most Bosnians as Bosniak or Islamic. “Mašić is a traditionally Croat name, while “Magazinović” is a name used primarily in Serbia, but is present in Croatia as well. “Srna Bajramović” is a compound of traditionally Slavic and Islamic family names.

107 leaders who claim to represent speciﬁc ethnic groups. The party denounces those leaders who “...use nationalist rhetoric of the threat to ‘us’ from ‘them’ and divert attention from the real problems of this society...”(Posavska Stranka 2010). The party cites its main concern as the economic underdevelopment of the Posavska region of Bosnia, which they say is because of the:

...administrative division into unnatural, ethnic and non-functional areas (the two entities and the Brčko District) as well as the incompetent, impenetrable and corrupt government that subordinated the public functions to personal rather than general interests....This and other factors has enabled the implementation of unfair and unethical privatization, increasing crime and corruption, legalization of unfair political and other privileges, the lack of a purposeful economic development and an increase in unemployment, a decline in social standards, the disappearance of the middle class and general impoverishment of citizens.

The party distances itself from the ethnic rhetoric of parties like the SDA and the HDZ, by laying out a doctrine of regionalism rather than the socialism and populism of the SDP. It rejects the arbitrary territorial divisions of the country into entity and cantons, but argues instead for a regional decentralization and local governance which it maintains would be more eﬃcient and less corrupt. On the other side of the country, in Una-Sana, Fikret Abdić’s DNZ also uses regionalism as an alternative to ethnic mobilization. Using his own personal base of support, and his reputation as a prominent Bosniak with strong ties to Serb forces during the war, Abdić portrayed the DNZ as a party based on a “Western Bosnian” identity independent of ethnicity, nationality, or religion. In a statement released prior to the 2008 election, the DNZ described itself as an opposition to the ethno-nationalism of the dominant parties: “The Democratic People’s Union of Bosnia and Herzegovina originated in Velika Kladuša in 1993 as a result of disagreement with the SDA’s policies and its leadership in Sarajevo who led Bosnia into total war and its citizens to exodus and pain....this newly formed party calls on all those who see nationalism, religious radicalism, and intolerance as a great danger for the isolated regions of Western Bosnia, and for Bosnia and its citizens (DNZ BiH 2008).”

108 Another subset of parties in the Bosnian context are remarkable for the way they completely ignore the ethnic cleavages present in Bosnian society. One such example is the SzBiH of Haris Silajdžić discussed above. Silajdžić is undeniably an important leader of the Bosniak community. He was a former member of the SDA, leaving in 1996 after conflicts with the party leadership (Šedo 2010, p. 88). He was part of the Bosniak delegation to Dayton during the civil war, and he has previously been elected to the Bosniak seat of the Bosnian presidency. He is un- apologetically in favor of a unitary Bosnian state—the exact position advocated by the Bosniak military and political leadership during the civil war, and the solution to the war that was most unacceptable to Croats and Serbs. Silajdžić is a polarizing figure, having stated to western film crews in 1995 that he “enjoyed” watching the NATO bombing of Bosnian Serbs, since Serbs would “finally feel what it means to be targeted; to be defenseless. And they deserved it.” (The Death of Yugoslavia 1995) Language like this often has the feel of an ethnic appeal: it clearly has a narrow audience, since claiming that Serbs “deserved” to be bombed would probably be much better received among ethnic Bosniak audiences than Serb audiences. Nevertheless, the SzBiH official positions and campaign statements often go out of their way to avoid making the explicit ethnic appeals that we see among other parties. The very first sentence of the SzBiH Program Statement claims, “[p]eople are by nature made different, but should in society and in law be equal (SzBiH 2015, p. 1).” Seeming to accept the diversity inherent in Bosnian society, the SzBiH nevertheless maintains that ethnic divisions should not be the basis of institutional design or policy making in Bosnia. The party further makes this clear in more direct statements: “It is [the principles of equality] that make it possible for the citizens and peoples of Bosnia and Herzegovina—historically a society of diverse faiths, ethnicities, and cultural traditions—to survive, accepting unity in differences and evolving in reciprocity, through practice and history (SzBiH 2015, p. 1).” The SzBiH does not overtly identify itself with a specific ethnic group. But unlike the socialist, populist, or regionalist parties described above, the SzBiH does not overtly attack the ethnic organization of the state as corrupt or inefficient. In fact, the SzBiH justifies all of its policy proposals in a way that largely ignores ethnic identity. On the subject of constitutional and administrative reform, the SzBiH seems close to the position of the Croat nationalist parties:

109 the constitution should be reformed and greater autonomy given to harmonized, local units. But this is seen as desirable because it will conform to European standards of eﬃciency and human rights, not because of its impact on any single ethnic group:

Bosnia and Herzegovina through the process of constitutional changes should ensure the full harmonization of the Constitution of BiH with basic human rights guaranteed by the international conventions and human rights chapters and eliminate from the existing Constitution both ethnic and territorial discrimination. Pro- tection of the rights to citizenship should be incorporated into the new constitution, as well as the right to returning property that is subject to the denationalization and restitution process. The new constitution should also provide a mechanism that will allow the rationalization of public spending through the creation of greater eﬃciency, e.g. increased capacity for the state, the revision of the provisions on education and public sector salaries, and the harmonization of this spending with the economic growth. The BiH Constitution should establish a single police structure of the BiH through the implementation of three principles of the European Commission, as a commitment undertaken by Bosnia and Herzegovina through the signing of the Stabilization and Association Agreement with the EU.(SzBiH 2015, p. 3)

On the subject of wartime reparations, the SzBiH claims:

We must provide all the elements necessary for the immediate return of over 150,000 refugees and displaced persons who have expressed their willingness to return to their homes. It is also necessary to provide the preconditions for a sustainable return of all those who have already returned to their homes or who will return. It is most important that the internally displaced citizens of our country who are accommodated in the centers are allowed as soon as possible to rebuild their homes and provide conditions for their undisturbed return. It is of utmost importance to maintain ties with the citizens of Bosnia and Herzegovina who live around the world who left their homeland mainly due to aggression, ethnic cleansing, geno-

110 cide and overall war destruction and demographic propagation of the population. (SzBiH 2015, p. 7)

Internationally, these positions have been warmly received. Silajdžić himself has often spoken in the west at prestigious universities and important policy research centers—an honor not shared by many participants in the Yugoslav civil wars. His support of efficient government, human rights, and refugee return often presents in Europe as forward-thinking, progressive, and democratic. But within Bosnia, the SzBiH’s appeal is limited almost entirely to Bosniaks. The extreme importance that the SzBiH places on homogenizing government structures and facilitating refugee return is seen by many as an open attack on the Serb entity, a sentiment Silajdžić has not dispelled by his referring to the Republika Srpska as “the genocide entity” while campaigning for office (Knezevic 2016). Croats often perceive the positions of the SzBiH as designed to limit the institutional protections that prevent small groups from being overrun and dominated by demographically larger groups—a position that favors Bosniaks over others. Thus while the SzBiH does not make overt ethnic appeals, the policies they advance seem designed to appeal to a single ethnic group. Similar issues are present with the People’s Party for Work and Betterment.24 Founded in 2001, the NSRzB presents itself as a multi-ethnic party, committed to civic identities and representation for all. The party’s founding statute declares that it will, “bring together all the peoples and citizens of Bosnia and Herzegovina of the same political attitudes and orientations. The NSRzB will represent the interests of all citizens of Bosnia and Herzegovina and shape public and political life on based on democratic principles in accordance with Western European standards (NSRzB 2008).” The party’s official platform and policy statements also frequently contain language attacking the ethnic organization of the state: “Every human is subject to mistake and guilt. Human imperfection and ignorance can create borders in politics. If we are aware of these borders, we can avoid dogmatism and totalitarian understandings of politics, and can create the possibility for reconciliation” (NSRzB 2004, p. 10). Yet there are convincing reasons to believe that the NSRzB is a Croat party. The party chairman, Mladen Ivanković-Lijanović is himself an ethnic Croat, and has run for president only

24“Narodna Stranka Radom za Boljitak”, NSRzB

111 in the Croat seat of the Bosnian state presidency The NSRzB is run primarily from the town of Široki Brijeg, in the overwhelmingly Croat canton of West Herzegovina. Historically, Široki Brijeg has been strongly associated with Croat nationalism. This reputation was so strong that the communist government actually renamed the town in order to undermine what they feared may have been a nationalist rallying point during the Yugoslav period.25 The party cannot really claim to be state-wide in scope, and attracts supporters mostly in the regions in the south-west of the country where Croats are concentrated. Moreover, the party’s multiethnic rhetoric is supported by a commitment to decentralization. As the party writes in their oﬃcial Statement of Principles:

Individuals and small communities are the units closest to daily life, and should be given the responsibility for solving their own problems. These smaller units should absorb the responsibilities of the entities and the state, which will be freed to focus on the planning and implementation of only those tasks requiring state authority. It is necessary to review the existing division of capacities and to relieve the state of those tasks that local institutions can do better.

We stand for the view that an individual should demand from the community only what he cannot do on his own.

There must be a clear separation of powers which extends from the local to the international level. The European Union’s principle of federalism, where autonomous municipalities and regions come together into states and larger communities is our guiding principle. (NSRzB 2004, p. 14-5)

This support of decentralizing is not, strictly speaking, an ethnically polarizing principle. Many countries, including the United States and Germany which are less ethnically diverse have federal systems and debate on the appropriate allocation of power between local and higher authorities. But the plea for decentralization in Bosnia is much more popular among Croats than it is among Bosniaks. Given their smaller size relative to the Bosniak population, a weak central government was a key demand of the Croat forces during the war, and centralization remains

25The locals returned the original name in the lead-up to the civil war (Hall 1994, p.49).

112 unpopular among ethnic Croat voters who frequently fear being overwhelmed by the larger Bosniak community. The fact that the party wins most of its support from Croats in Croat areas on a platform of decentralization which tends to be popular among ethnic Croat voters creates an impression at odds with the party’s presentation of itself as a multi-ethnic, entirely program- matic organization. The party is often discussed in the Bosnian media as Croat party which competes only with the HDZ and other Croat nationalist parties.

4.5 Classifying Bosnian Political Parties

From the descriptive analysis above, I argue that the best way to understand Bosnian parties is with a qualitative three-part typology: ethnic, anti-ethnic, and ethnically ambiguous. Such a categorization system captures the most important variation in political party strategies, and facilitates the analysis of Bosnian voting behavior in the next chapter. While all parties fall into one of these three categories, only one type meets the standards of an ethnic party as described in chapter 2. First, I outline this classification system. Then, I respond to potential concerns about why the differences between ethnic parties and other types of parties is important to understand both in a general theory of ethnic voting, and the Bosnian case specifically.

4.5.1 A Three-Part Typology

Many parties in Bosnia meet the definition of an ethnic party. Parties such as the SDA, HDZ, and SDS campaign almost exclusively on ethnic communal interests. In their choice of party name, logo, rhetoric and campaign strategy, they consistently strive to ensure that voters associate their party with a specific ethnic group, and provide a clear understand that the party will favor coethnics at the expense of ethnic outsiders. The parties even decry inter-ethnic cooperation, using ethnically divisive language to describe policy proposals that would likely have broad multi-ethnic support. For these parties, policy positions may be vague or fickle, but on the question of who the party will represent, they are nothing if not steadfast. While identifying an ethnic party is relatively straightforward, it is less clear what to make of those parties that do not meet the definition. For that, I propose that the remaining parties

113 should be subdivided into two categories. This ultimately creates a three-party typology into which all Bosnian can parties fall. Some parties may be thought of as anti-ethnic. These parties openly attack ethnicity as the basis of legitimate political mobilization. They clearly communicate to voters that if elected, no citizen should expect to benefit because of their ethnic identity at the expense of citizens from other groups. Parties like the SDP and the DF that insist on a civic identity and attempt to communicate to voters that they will show no favoritism to any specific group fall into this group. Just as an ethnic party speaks directly to a specific ethnic group, an anti-ethnic party vocally disavows any connection to a single ethnic community. Anti-ethnic parties acknowledge that Bosnia is an ethnically divided country and concede the possibility that political blocs may form on the basis of identity, but argues that such an outcome would be normatively undesirable. In doing so, the party makes it clear that it is rejecting ethnic identities as the basis of policy- making. Those parties which do not fit into either of the other two categories can be thought of as ethnically ambiguous. These are parties which neither accept nor reject ethnicity as the basis of political action. They do not seek to build a political base of support using ethnic identities, nor do they actively seek to create a multi-ethnic group of supporters. These parties do not take positions on ethnic issues. They treat the country’s ethnic divide as unimportant, or ignore it completely. Parties like the SzBiH and the NSRzB may get most of their support from a single ethnic group, but this is not necessarily because they refuse or discourage support from other ethnic groups. The party may advocate policies that only a single group would find appealing, but if the policy is advocated on its own merits, and not because it will advantage a single group, then the party can still be coded as ethnically ambiguous. The typology I employ here is important because it facilitates an explanation demanded by the research question. It is minimally descriptive, dividing Bosnian political parties into categories based only on the most pertinent category for the question at hand: the use of ethnic identities. In keeping with the qualitatively different nature the different categories, this system is nominal rather than ordinal or absolute (Collier, LaPorte & Seawright 2012). The three types of parties are descriptive categories which cannot be quantitatively compared to each other.

114 Type Characteristics Examples Claims to Represent Ethnic Ethnic Party SDS, SDA, HDZ Community Rejects Ethnic Favoritism; Anti-Ethnic Party Cultivates Multi-Ethnic Base SDP, DF of Support Ignores Ethnic Division of Ethnically Ambigu- Country; Does not use Eth- SzBiH, NSRzB ous Party nicity as Basis of Inclusion or Exclusion

Table 4.1: Bosnian Party Typology

They cannot be arranged spatially: an ethnically ambiguous party is not somehow “less” ethnic than an ethnic party or “more” ethnic than an anti-ethnic party. Instead, the three are fundamentally and qualitatively different categories used to describe how a party employs ethnicity as a political asset. The three categories are mutually exclusive and collectively exhaustive, facilitating the study of Bosnian political activity by reducing complexity, consolidating similarities, and differentiating contrasts (Bailey 1994, chap. 1). On the subject of ethnicity, all Bosnian political parties must do one of three things: adopt an ethnic identity, reject ethnic identity, or ignore the question altogether. The typology described here allows for variation in relationship between ethnicity and voting that can help provide explanations for voting behavior. It allows for each vote cast in Bosnia to be coded as one of these three categories, and facilitates the quantitative analysis of qualitatively different behavioral observations. In this, I follow King, Keohane & Verba (1994, chap. 2) in using a typology to provide conceptual clarity in quantitative study. In the next chapter I use this typology in a statistical analysis of Bosnian voting behavior.

4.5.2 Can a Bosnian candidate really be “non-ethnic?”

It may be objected that the typology I propose overestimates the diﬀerences between parties, because all parties in Bosnia are in some way “ethnic.” Intuitively, any form of political activity that is not in some way inﬂuenced by ethnicity seems hard to imagine in Bosnia. Ethnic identities are entrenched in virtually every aspect of life, and their social salience has been reinforced

115 by recent violence. Institutions and social convention segregate groups in a way that minimizes meaningful inter-ethnic activity. Given the extreme sensitivity of Bosnian voters and politicians to ethnic issues, and the highly visible way in ethnicity is communicated through family names and speech patterns, it may seem that a non-ethnic or even an ethnically ambiguous party is simply impossible. These concerns merit serious discussion, but ultimately, ignoring the possibility of parties which are not necessarily ethnic in nature obscures important insight to be gained from a study of ethnicity in Bosnian politics. One valid concern is that “non-ethnic” appeals cannot be equally appealing to all ethnic groups. It is true that during the civil war, bargaining positions and strategic objectives were closely linked to each group’s social status within Bosnia: Serb forces advocated for secession, Croats advocated for federalism, and the Bosniaks advocated for centralization. These positions are still among the main policy proposals advocated by ethnic parties in each faction, and it is plausible that these positions are the still logical consequences of each groups’ position within Bosnia. As the largest group, Bosniaks would likely dominate a centralized government that had no explicit ethnic protections. Given their small size and territorial concentration, Croats would gain political power and autonomy under a further decentralized state. If policy preference is a function of ethnicity, then a policy proposal is no different from an ethnic appeal, and treating the two as separate is inappropriate. It may also be argued that ethnic identities in most cases cannot be hidden in Bosnia. The moment a person declares herself a candidate she identifies with a specific identity group, and that identification will inherently influence people’s thinking in a social context where ethnic divisions are pervasive. A Bosnian voter who sees a party headed by a chairman with an obviously Turkish or Islamic name can readily assume that the candidate is a Bosniak, and may also infer that the party will support the Bosniak community if elected. If this is the case, then all the party rhetoric about a multiethnic state and civic nationalism is essentially empty talk. It has also been argued in other contexts that appeals to non-ethnic identities can themselves be a way to reinforce the dominance of one ethnic group over another. Sociologists and critical theorists of race relations world have often argued that the “invisibility” of ethnic iden-

116 tities is in fact a system which reinforces pre-existing privileges.26 In this sense, while the SDP may advocate a country where ethnic divisions are of no import, they may actually be supporting the preservation of a specific system of ethnic relations linked to Yugoslav communism. Their pleas for a strong centralized state which provides generous social welfare benefits could be interpreted as proposal to return to the era of Titoist administration, when most powerful Bosnian politicians were ethnic Bosniaks and Serbs tied closely to Belgrade. A proposal for an increased emphasis on human rights and accountability may sound “ethnically blind” to western observers, but has enormous implications on ethnic group relations given Bosnia’s history. Since the current Bosnian institutional arrangement linking political authority, territory, and ethnicity is largely possible because of genocide, ethnic cleansing, and targeted violence against civilians, focusing on accountability for human rights violations may be seen as an attack against the existing order which grants equal status to all three ethnic groups. These appeals are particularly hostile to Serbs, as the Republika Srpska is set apart from the other two groups largely as the result of military conquest and 1990’s era genocide. Even appeals to local autonomy and federal decentralization may be seen as a way to reinforce existing norms which benefit certain groups over others. Croats, for instance, may find local autonomy appealing because it empowers small groups concentrated in specific areas (i.e., Croats), but not extremely small groups scattered through the country (i.e., Jews, Roma, and Turks). Yet empirically, it remains undeniable that variation in the use of ethnic identities among political parties exists, and that this variation is systematic. Given the high degree of consistency in the use or rejection of ethnic identification among specific actors, this variation is best explained by parties and candidates making conscious decisions in pursuit of political goals. Po- litical actors cannot easily control or change their own ethnic identities, but they can change how they present those ethnic identities to voters. It is clear from the summary above that Bosnian politicians expend enormous resources to accentuate or downplay their ethnic identities. Some parties—those I describe as ethnic—talk about nothing but ethnicity, even when it doesn’t seem strictly necessarily. On issues like the environment, security, or EU accession, ethnic parties

26See Carby (1992), Roman (1993), Dyer (1997), and Dalton (2008), among others. While much of this literature is focused on racial identities, recall from Chapter 2 that the deﬁnition of “ethnicity” in this dissertation is any identity associated with descent. Under this conceptualization, racial identities are ethnic identities.

117 feel the need to differentiate themselves from parties representing other ethnic groups, even when the two parties are in agreement. This suggests that it is so important to their strategy to provide an ethnically-based justification for their actions that they are willing to antagonize other ethnic groups (and potential coalition partners) just to ensure their own ethnic brand is strong in the minds of voters. Non-ethnic parties, likewise, go out of their way to avoid being perceived as affiliated with a single ethnic group. Cognizant that voters may be inferring an ethnic attachment from the names of the candidates or party leadership, the SDP purposefully puts together obviously multi-ethnic party lists, to prevent the impression of Bosniak dominance in the minds of voters. The NSRzB goes out of its way to put voter platforms in all three languages (despite complete mutual intelligibility) on its web page, to bolster its credibility as a party that shows no favor to any single ethnic group. From the outside, it may appear that ethnicity is the all-consuming issue in Bosnian politics. But to ignore this variation risks ignoring an important variable that links ethnic diversity to democratic outcomes. Rejecting the differences between the different types of parties further risks over-reliance on an essentialist view of ethnic identity, and a rejection of human agency. It is true that Bosnia is ethnically divided, that Bosnia has a recent history of extreme ethnic violence, and that ethnicity in highly polarizing in Bosnia. But to claim that because of these underlying conditions all political action is explainable by ethnicity assumes that there is something inherent in ethnic identities that will automatically be more politically relevant than all other identities, preferences, or social cleavages. Such an interpretation implicitly endorses the primordialist notion that ethnic identities are “givens” in the Bosnian context. This is not consistent with recent Bosnian history. As explained in Section 1 above, the very existence of the largest ethnic group in Bosnia was still debated as recently as the 1970’s. It has long been observed that ethnic identities can appear to be “natural” for a variety of reasons.27 Nevertheless, accepting ethnicity as a natural form of self-identification or inherently politically relevant risks ignoring the processes which created the identity and gave it social salience in the first place. Ethnic identities may be slow to change, and individuals may have relatively little capacity to change the political significance of ethnicity in the short term, but that is not to say they have no agency at all. Politicians

27See Barth (1969), Bentley (1987), Cornell (1996), Hirschfeld (1998), Gil-White (2001), Wimmer (2008), and Brubaker (2009).

118 can choose how to present themselves to voters, and whether they will emphasize their ethnicity or not. Ignoring this variation begs the question: what does non-ethnic politics in an ethnically divided society look like? Since every person has an ethnic identity, should we assume that every action is a manifestation of their ethnic identity? If we deny political actors the possibility of political action independent of their identities, we essentially assume the inevitability of ethnic division in any diverse democracy. The assumption would make it impossible to participate in any form of political community that transcends ethnicity, condemning any non-homogeneous society to what Howard (2012) calls “the ethnocracy trap.” Ethnic political mobilization, even in countries as polarized as Bosnia, must not be an assumption, but rather a phenomenon to be scrutinized and subjected to falsiﬁable empirical testing. The typology proposed here allows for variation in ethnic mobilization, and in the next chapter uses this classiﬁcation system to empirically measure the degree of success for each type of appeal.

4.6 Conclusion

The three main ethnic groups in Bosnia are all descended from a common ancestor, and share much in history, culture, language, and tradition. While the distinction between these groups is currently institutionalized by the Bosnian political system, this has not always been the case. Even within living memory, Bosnia was governed by a socialist regime that sought and often succeeded in integrating the three groups into a cohesive multiethnic society. While the ethnic civil war of the mid-1990’s devastated Bosnian society, it also produced a new set of political institutions designed to facilitate electoral competition and democratic governance. Under this system, political entrepreneurs seeking to win elections use ethnicity in varying ways. Some actively invest in their ethnicity, making exclusively ethnic appeals to an exclusively coethnic constituency. Others are open to multi-ethnic bases of support, either vocally denouncing ethnic segregation or simply ignoring it entirely to frame their proposals in more broadly acceptable terms. The result is that Bosnian voters ﬁnd themselves living in an ethnically polarized society,

119 where ethnic parties compete directly with non-ethnic parties in the same elections. In chapter 2, I argued that voters in such a situation should systematically vary in their willingness to support ethnic parties depending on institutional and demographic factors. In the next chapter, I use the framework outlined here to test these theories through a quantitative analysis of Bosnian voting behavior.

120 5

Ethnic Voting in Bosnia and Herzegovina

In the previous chapter, I outlined the qualitatively different types of appeals made by Bosnian politicians to voters during elections. I argued that all party positions can be categorized by a three-part typology: explicitly ethnic, explicitly non-ethnic, or ethnically ambiguous. In this chapter, I use that typology to test predictions generated by the theory outlined in Chapter 2. There, I argued that demographics directly influence voters’ assessments of the benefits to ethnic policy platforms as well as the barriers to entry for ethnic group representation. In many contexts, it is difficult to disentangle the effects of demography on these two dimensions, because it is usually not feasible to observe variation in the demographic circumstances while holding other factors constant. Fortunately, such variation is present in the Bosnian context. The constitutional structure of post-Yugoslav Bosnia and Herzegovina requires voters to elect leaders to multiple bodies of government simultaneously under varying demographic circumstances. Elections to the political entities under study here are coterminous, which holds constant voters’ political preferences, turnout rates, and international and economic political contexts, while varying demographic conditions between levels of government. In this chapter, I exploit that variation. Mapping the qualitative differences in the ways which parties use ethnicity on to quantitative voting data, I show that Bosnian voters are quite willing to vote for ethnic and non-ethnic parties even in the context of a single election. Split- ticket voting between ethnic and non-ethnic parties is extremely common in Bosnia, despite the fundamental contradiction in the logics these parties use to justify their positions. I argue

121 that this behavior is explainable by the variation in demographics between the different levels of government in Bosnia. When ethnic representation is most instrumentally useful, voters support ethnic appeals. When it is less useful, voters support parties that do not campaign as champions of ethnic groups, favoring ideological or non-ethnic populist parties. In the first section of this chapter, I explain why an analysis of split-ticket voting in the Federation of Bosnia and Herzegovina is a useful tool to identify the effect of group size on the willingness of voters to support an ethnic party. In the next section I describe the novel data set used to conduct this analysis. The third section shows the results of the analysis, and argues that split-ticket voting in Bosnia reflects systemic patterns of ethnic party support consistent with the predictions outlined by the theoretical chapter. Section four presents the observable implications of the general pattern at a more micro level, examining a single Bosnian municipality and the specific parties which ran for office at both levels there. Section five provides a brief overview of the same pattern observed over time, and section six concludes.

5.1 Split-Ticket Voting and Ethnicity

Within the Federation of Bosnia and Herzegovina, voters can find themselves citizens of a municipality, a canton, an entity, and a state. Each of those polities has an independently elected legislative assembly, meaning that a voter must vote to fill four separate legislatures. Voters may find themselves in drastically different demographic circumstances at each level. The majority of Croats live in Croat-majority cantons, and the majority of Bosniaks live in Bosniak-majority cantons. Territorial ethnic homogeneity is not perfect, though. As it is essentially impossible to draw a territorial boundary that perfectly includes a single ethnic group while excluding all others, a significant number of people found themselves on the “wrong” side of a cantonal border at the end of the war, living in cantons that are legally defined to be representative of another ethnic group. Moreover, while the boundaries between cantons and entities have not changed since the signing of the peace accords, internal migration and refugee returns have served to increase the ethnic diversity of many cities and towns. As a result, while cantons can be said to be much more ethnically homogeneous than the country as a whole, they nevertheless retain

122 pockets of ethnic minority voters. Cantonal Assemblies and the State House of Representatives are both parliamentary bodies, elected by an open-list PR ballot. Yet the barriers to entry for would-be politicians are clearly much higher at the state level than they are at the cantonal level. The state-level House of Representatives has 42 seats, elected from both the Federation and the Republika Srpska. Elec- toral districts, whose borders overlap with cantonal boundaries in the Federation, elect multiple members, the lowest district magnitude being three and the highest being six. Mathematically,

1 this means that a party or candidate is guaranteed a seat in the parliament if they can win 6 of the vote share in the least restrictive district. Compare this to the cantonal assemblies, which are also selected from an open-list PR ballot using the canton as single electoral district. The smallest Cantonal Assembly in BiH is Posavina, with 21 seats, and the largest (a three-way tie between Sarajevo, Tuzla, and Zenica-Doboj) have 35 seats. This means that even in the most

1 restrictive case, a candidate is guaranteed a seat if they earn 21 of the total vote share. Can- tonal Assemblies have a much lower barrier to entry for smaller parties, and consistent with the literature on party systems and electoral laws, we should expect a much more fragmented party system at the cantonal level, with greater ease for cantonal-level ethnic minorities to gain representation. A summary of the differences between the Cantonal Assemblies and the State House of Representatives is shown in Table 5.1. In keeping with the consociational model, the Bosnian constitution is designed to put as many people as possible in situations of low barriers to entry for ethnic parties. The arrangement actually creates divergent patterns of benefits and ease of access across ethnic groups at different levels of government. Consider the example of a Bosniak who lives in majority Bosniak canton. When voting at the cantonal level, the benefits of ethnic representation are relatively low. Coethnics are likely to dominate both private and public sectors simply by virtue of their demographic advantage. The Bosnian language would likely dominate because most of the canton belongs to the same linguistic community, and would thus need relatively little protection. The possibility of diverting resources specifically to other Bosniaks–i.e., taxing ethnic outsiders to the benefit of fellow Bosniaks–yields a small pay-off, because the target population that could be expropriated is relatively small. In short, ethnic representation is unprofitable, and ideology,

123 Cantonal Assemblies House of Representatives 1 3 (Either Bosniaks or (Bosniaks, Croats, and Constituent People Croats) Serbs) The entire multiethnic 1 of 10 ethnically homo- Territory Governed state of Bosnia and geneous Cantons Herzegovina Polity Population <500,000 4 Million District Magnitude High (21-35) Low (3-6) Low for Cantonal Ma- High for Cantonal Ma- Value of Ethnic Repre- jorities; High for Can- jorities; High for Can- sentation tonal Minorities tonal Minorities High for Cantonal Ma- High for Cantonal Ma- Ease of Access for Eth- jorities; High for Can- jorities; Low for Can- nic Parties tonal Minorities tonal Minorities

Table 5.1: Institutional and Demographic Differences between Cantonal Assemblies and House of Representatives class, or some other social cleavage will likely pay more dividends as the basis of representation. The ease of access for Bosniak ethnic parties in this scenario, however, is extremely high. Since institutions are designed to create an overwhelmingly large majority group in ethnically homogeneous cantons, voters can be sure that ethnic representation is a completely viable option with hardly any electoral obstacles. The electoral institutions compliment the demographic circumstances to give ethnic parties appealing to the majority group virtually no risk of coordination failure or vote wasting. With an extremely high district magnitude for elections to the cantonal assemblies, parties appealing to the majority group are almost certain to be able to win. In this case, for those in the majority group the homogeneous nature of the cantons presents a typical majority group outcome: a clear path to ethnic representation, but relatively little to gain. The circumstances are different, however, for elections to the state-level parliament. Majority-group voters can be reasonably certain that ethnic representation is still possible, since they remain the overwhelming majority in their district. Even though the district magnitude at the state level is substantially lower, creating increased obstacles to representation for smaller political groups, ethnic majorities can remain confident in their ability to clear those hurdles provided an organized mobilization effort. The potential benefits of ethnic representation, though, have changed dramatically. Now, the body to which voters are electing representatives is gov-

124 erning a fundamentally different body of citizens. This authority is charged with setting policies governing the entire country, which is much more ethnically diverse. The dominance of the local-level majority group cannot be assumed at this point, since the other two groups are also electing leaders to parliament. Numerical dominance is no longer sufficient to ensure protections for language and culture, and electing ethnic champions may have increased appeal to voters concerned about discrimination or oppression. Moreover a platform of diverting resources from ethnic outsiders to benefit coethnics has a much larger target population to tax, increasing the potential rewards to each voter of electing an ethnic party to office. In short, a voter living inside a canton dominated by coethnics has varying incentives to support an ethnic party at the cantonal level and the state level. Ethnic parties are viable at both levels, but have much more to offer at the state level than the cantonal level. This voter should therefore be more likely to support an ethnic party in the House of Representatives than in the Cantonal Assembly. For a minority group member, however, i.e., a voter from a group who is not a dominant majority within her canton, the situation is reversed. Whereas a majority group member has relatively little to gain through ethnic representation at the canton level, a minority group member has much. As a minority, she is likely to live in an environment where her language is spoken by relatively few, and could be dominated by the larger numbers of the majority group without explicit legal protection or media subsidies. A preference for hiring coethnics or blatant discrimination among the majority community could make it harder for her to gain access to jobs or state resources due to her ethnic outsider status. In a country like Bosnia, where ethnic violence is a very real memory for much of the population, security issues could also be extreme. The dominance of the police, justice system, or security services by local majority groups may mean an increased risk of intimidation or expropriation. In addition to these risks which a voter would like to avoid, there are also potential rewards to be gained through ethnic representation. Financial transfers to minority group members would effectively tax a very large proportion of the population to redistribute to a small portion of the population, providing a potential windfall of resources. For all of these reasons, ethnic representation is likely to be very appealing to a minority group voter in the canton in which she lives, resulting in a very high level of benefits at the cantonal level. Since the state is also

125 highly diverse, all these same arguments would apply to the state House of Representatives, meaning that level of benefits to be gained by ethnic representation is quite high at the local and highest levels of government for minority groups. Even though the potential benefits of ethnic representation are high at both levels of government for cantonal-level ethnic minority groups, the viability of such ethnic representation varies drastically between the two levels. Ironically, despite the intent of the Bosnian constitution’s designers, ethnic minority groups are most viable at the cantonal level where they are ethnic outsiders—not the state level. Since Cantonal Assemblies are composed of dozens of seats, and use the canton as a single electoral district, less than three percent of the vote may be enough to guarantee a legislative seat at the Cantonal level. This drastically reduces the risk of vote wasting or coordination failure that ethnic minority voters may face when voting, compared to the state level. At that higher level of governance, district magnitude is much lower, where more than five times the vote share would be required to be guaranteed a seat in the state parliament. As such, the ease of access for smaller groups is substantially higher at the Cantonal level than it is at the State level due to reduced barriers to entry. This produces several hypotheses of Bosnian voter behavior. First of all, against theories of ethnic heuristics, or preference derived from ethnic attachment, Bosnian citizens should vary in their willingness to support ethnic parties based on the political unit whose offices they are electing. Since the potential gains of ethnic representation and the likelihood of actually obtaining it are not the same at both levels of government, we should not expect to individual voters to respond to ethnic appeals made by parties in the same way in both contexts. Secondly, the variation in voter willingness to support ethnic parties should be a function of their local-level demographic status. Cantonal majority group should be more likely to support non-ethnic parties at the cantonal level, but switch their vote to ethnic parties at the state level. This also means that an increase in ethnic voting at the state level should be directed to cantonal-level majority groups (and not cantonal-level minority groups). Cantonal minority groups should exhibit the opposite behavior, supporting ethnic parties at the cantonal level, but non-ethnic parties at the state level. A summary of these predictions is listed in Table 5.2.

126 Vote share for ethnic and non-ethnic parties will vary between cantonal 5.1 and state-level elections Ethnic voting in elections for the House of Representatives will decline 5.2 in proportion to those votes going to ethnic minority parties for the Can- tonal Assembly Ethnic voting in elections for the House of Representatives will increase 5.30 in proportion to those votes going to non-ethnically identiﬁed parties for the Cantonal Assembly Increases in ethnic voting in the House of Representatives Elections will 5.31 be driven by increased support for parties representing local ethnic majorities, not minorities

Table 5.2: Hypotheses

5.2 Data Collection and Sources

The fact that Bosnian institutions have voters electing officials simultaneously under different institutional environments presents a unique opportunity to test a theory of ethnic voting based on the desirability and feasibility of ethnic representation. Elections for cantonal and state-level parliaments are held concurrently, and the voter’s perception of pressing policy issues or preferences are not likely to change between local and national-level elections. Under permissive electoral laws and a highly fragmented political party system, the same political parties compete at both subnational and national-levels for political office. Under Bosnia’s ethno-federal constitution the cantonal governments and the state government often have overlapping authority, suggesting that voters are likely to evaluate candidates by the same criteria at both levels of government. To exploit this opportunity, I construct a novel data set that consists of two parts: a qualitative coding of which parties in Bosnia make ethnic appeals and which do not, and precinct-level voting data collected directly from the Bosnian Election Commission. The data set covers all elections in Bosnia 2006-2014. The time period under study starts more than ten years after the end of the Yugoslav Civil War, at which point international election observers and NGOs had certified that elections were free and democratic.

127 5.2.1 Coding Parties

Party registration rolls from the period under study contained 149 different political parties, pre-electoral coalitions, and individual candidates. Many of these entries were duplicates or alternate abbreviations of parties and candidates listed elsewhere. After consolidating the list of parties registered with the Bosnian Election Commission to eliminate these discrepancies, I was left with 91 unique parties, coalitions, and individual candidates. I then coded all political parties as either ethnic parties representing Bosniaks, ethnic parties representing Croats, ethnic parties representing Serbs, or non-ethnic parties. This was done based on party platforms, party information, candidate declarations, and mission statements downloaded from publicly available political party websites. Of the 91 registered electoral choices, 20 of them were independent candidates or parties so small that they had no trace in the public record that I could find. I therefore drop these votes from the dataset, treating them as if the people who had voted for these candidates had simply abstained. Since these parties are extremely small and had very low levels of support, it is not expected that they should systematically bias the analysis in any direction. In total, the coding scheme covers 98.48% of all votes cast in all three elections. In coding parties, I rely on the definition of “ethnic party” outlined in Chapter 2. For a party to be classified as ethnic in this dataset, it must clearly identify itself with a specific ethnic group. It must also exclude other ethnic groups from their agenda, making it obvious who the “outsiders” are who will not benefit from the policies enacted by the party once in power. The party must also make its ethnic identity central to its brand, with ethnically-defined group interests as the central mission of the party’s political agenda. The standard employed is whether or not a reasonably intelligent voter would be able to clearly and quickly identify which ethnic group a party claimed to support using information made readily available by the party itself. While the criteria applied may be considered somewhat subjective, the nature of campaigning in the Bosnian context provides a certain degree of clarity. Because of the political party activities outlined in Chapter 4, there are relatively few controversial coding decisions. Many parties rely so heavily on ethnic cues in campaigning to voters that the name of the ethnic group is also the name of the party, and national or ethnic symbols are frequently used as party logos. For these parties, which ethnic group the party represents is abundantly clear from immediate

128 observation, especially to Bosnian voters who are often extremely sensitive to ethnic identification. Other parties explicitly attack ethnic parties as corrupt and denounce ethnic divisions as immoral and illegitimate, making their own non-ethnic identity extremely obvious. The potential difficulties in coding stem from the category defined as “ethnically ambiguous” in Chapter 5. These are the parties which completely ignore ethnic issues in their campaigning, neither explicitly identifying with a single ethnic group, nor publicly declaring their advocacy of interests on explicitly non-ethnic grounds. For the purposes of this data set, these parties are classified as non-ethnic. This is a deliberate decision, and one that is consistent with the approach and aims of this dissertation. First, the definition of “ethnic party” used in this dissertation requires explicit ethnic identification. This high standard is the most appropriate to explaining the choices people make in mobilizing along ethnic grounds during elections. The definition creates a measurement standard that captures the variation most important to this question: the choice of politicians to identify as an ethnic champion, and the choice of voters to support them. Nevertheless, this standard puts my coding decisions at odds with other definitions commonly used in the literature. Many parties in Bosnia meet the standard used in Horowitz (1985) in that virtually all of their voter support is derived from a single ethnic group, but nevertheless present themselves in non-ethnically defined terms. Many regional experts on Bosnia classify these parties as ethnic. Šedo (2010), for instance, classifies the Party for Bosnia and Herzegov- ina (SzBiH) as a Bosniak nationalist party, whereas I classify it as a non-ethnic party. There are admittedly convincing intuitive reasons to label the SzBiH as an ethnic party: its political support comes almost exclusively from Bosniak areas, and it was originally formed as a result of an internal rift within the front-running Bosniak nationalist party.1 But to classify the SzBiH as ethnic because of which ethnic group supports it assumes the primacy of ethnicity as a mechanism for vote choice, and thus cannot be used to analyze the relationship between ethnicity and voting. Coding parties such as the SzBiH as non-ethnic holds that voting for a party that explicitly advocates for the interests of a single ethnic community is categorically different than voting for a party that does not. As such, the coding system allows for the possibility that ethnic

1See discussion beginning on page 108.

129 identity is epiphenomenal to some other interest, and is more helpful in explaining why voters are mobilized along ethnic lines. The biggest concern with this standard is that Bosnian voters do not judge a party by their words, but by their supporters. It is entirely possible that voters do not form ideas of parties’ ethnic allegiances based not on what they claim they will do (or even what they actually end up doing) but by their base of supporters. This problem may be exacerbated by the Bosnian media, which often presents even ostensibly non-ethnic parties as if they catered to a single group. Politicians may have incentives to portray competitors as being beholden to ethnic groups as part of an electoral strategy, further compounding the problem. If this is the case, then the perceptions driving voting behavior are not adequately measured by the party campaign strategies, since voters assess parties by factors not directly under the party’s control. Given the high degree of salience that ethnicity has in Bosnian daily life, and the complete lack of stigma surrounding explicit ethnic appeals in Bosnia, I do not believe this is the case. Nevertheless, I cannot completely discount that these assessments may be the ways in which voters understand parties when making voting decisions. As such, I recode the parties classifying these ethnically ambiguous parties as ethnic parties in Appendix A as a robustness check.

5.2.2 Voting Data

Voting data was obtained directly from the Bosnian Election Commission using a series of automated web scraping scripts. Voting results were collected at the precinct-level for all three elections under study. Precincts in Bosnia are quite small, bringing them as close as possible to the individual-level. Pooling results from all three elections produces over 6,000 observations (after excluding the two mixed-ethnicity cantons), with mean and median precinct size of roughly 300 people. As precincts represent individual neighborhoods within municipalities, they tend to be much more ethnically homogeneous than the cantons and municipalities as a whole. Precincts are not perfectly homogeneous, though, and there is the potential that systematic differences between group members could bias the results. However, this bias would most likely be towards null effects, with minority and majority members canceling each other out. Moreover, the relatively high sample size of this study can increase confidence that any results

130 observed are the result of systematic behavior. While precincts are quite small, they remain amalgamations of individuals. This raises the possibility that any findings could be affected by ecological fallacy. One possible solution to this would be to rely on polling data of individuals, such as exit polls or voting intent data. Such an approach would have the advantage of bringing the unit of analysis in line with the level where the theorized mechanism operates. However, there are drawbacks to this approach. The most obvious is availability of data. Bosnian public opinion data is not always readily available, and many of the surveys which do exist do not always explicitly differentiate between cantonal and state-level categories among the same group of respondents. Survey respondents may not be the most reliable sources of data on their own behavior, as studies in other countries have demonstrated a general trend for under-reporting split-ticket voting in a variety of contexts. Burden & Kimball (2009) note that comparing self-reported survey data to actual election results in the United States shows that voters report higher levels of straight party ticket voting than are borne out by the actual electoral results. Burden (2009) shows that in Japan, voting for smaller niche parties goes under-reported in self-reported survey data. Since some parties may be more viable in either the Single Member District or Proportional Representation side of Japan’s mixed electoral system ballot, the under-reporting for small parties creates a systemic bias where substantially less ticket-splitting voting is self-reported than is actually observed. These concerns are especially potent in this context, as the Bosnian ballot is enormously complicated. In some elections, a Bosnian voter could be expected to vote for a municipal councilor, a mayor, a member of a cantonal parliament, a member of an entity parliament, and a president all at once. When asked to recount who they voter for at each election, a voter could be forgiven for confusing exactly who she voted for when. Moreover, the concerns of misreporting driven by conscious or unconscious social desirability bias may be especially acute here, as survey respondents would be in a position of having to admit that they voted for parties espousing two contradictory arguments in the same electoral contest. Voters may not be willing to admit to this behavior, and as such relying on self-reported data would introduce a strong possibility for bias towards null results.2 Using actual electoral data alleviates these concerns,

2Anecdotal evidence from my own ﬁeldwork suggests that the social desirability bias against admitting split- ticket voting may be quite strong in Bosnia. In many informal discussions with Bosnian academics, political ﬁgures,

131 since the behavior is observed directly.

5.3 Analysis

5.3.1 Descriptive statistics

Conventional wisdom holds that most voting is ethnic voting in Bosnia. While this is indeed the case, the degree to which this is true may have been overstated. Pooling cantonal and state- level elections over the three electoral cycles surveyed, only 51.75% of the vote went to parties explicitly identifying as ethnic.3 This focus on state-level aggregation though ignores important district-level variation. Figure 5.1 hints at this variation by displaying a box plot of the vote share going to ethnic parties in the entire data set. In this figure, the solid line in the box represents the median vote share going to ethnic parties at either the cantonal or state-level legislative elections among all precincts. The box represents the first and third quartiles, and the whiskers extending beyond the box represent the 1.5 times those values to cover the ethnic vote share of 3/4 of the precincts surrounding the median. Not surprisingly, the ranges covered by the second and third quartiles are rather large. This suggests a large degree of precinct-level voting heterogeneity, which is generally to be expected in most democracies, as people are not randomly assigned to voting precincts. Even at this level of aggregation, the median precinct shows greater ethnic voting at the state level than the cantonal level. While the median precinct supported ethnic votes in cantonal legislatures at the rate of 28.57%, this level rises to 36.69% at the state level. A C test of and everyday citizens, I mentioned that I was studying Bosnian elections, and had found that a substantial number of Bosnian voters engaged in split-ticket voting between ethnic and non-ethnic parties. No single person I discussed this with ever admitted to voting for different parties at different levels of government themselves, but offered up a variety of explanations as to why someone else might. I never revealed what the actual pattern was, but interestingly, people offered intuitive explanations that diverged in their prediction. Some thought that ethnic voting was more likely at the local level than at the national level, suggesting that some voters were fine with living in a multiethnic state, and respecting ethnic and cultural diversity at the national level, but disliked the idea of actually having to tolerate ethnic outsiders in their immediate vicinity. Others thought that ethnic voting would be more likely at the national level, suggesting that contact with ethnic outsiders in immediate circumstances would lead people to think that all of the out-group members they knew personally were “the good ones,” even if they distrusted the other groups as a whole in the abstract. No one seemed to offer a sympathetic or supportive explanation for this behavior, mostly seeing it as inconsistent or attributing it to ignorance or mercenary motives. While this evidence is completely anecdotal and no conclusions should be drawn from it, it is at the very least suggestive that a social desirability bias against split-ticket voting may skew results. 3If ethnically ambiguous parties are reclassified as ethnic parties, this number increases to 63.55%

132 Ethnic Voting in Bosnia and Herzegovina 1.00

0.75

0.50

0.25

0.00 Vote Share for Ethnic Parties Cantonal State Assemblies House of Representatives

Figure 5.1: Ethnic Voting in Bosnia at Cantonal and State Levels mean ethnic party support between Cantonal Assembly and House of Representative elections produces a change of 6.9%, with a C value of 29.004 (? < .001), suggesting a modest, but highly statistically significant tendency towards increased ethnic voting in House of Representatives elections over voting for Cantonal Assemblies. This indicates that a not insignificant number of Bosnian voters are varying in their support of ethnic parties between different levels of elections. Figure 5.2 disaggregates local-level majorities and minorities. The left hand side of this figure shows the vote share at both the cantonal and the state level going to parties who represent the cantonal-level majority, i.e., it combines the shares of parties claiming to represent Croats in Croat Cantons, and parties representing Bosniaks is Bosniak cantons. The right-hand side shows the vote share at both levels going to parties who represent the cantonal level minority. This include parties representing Bosniaks in Croat cantons, or Croats in Bosniak cantons.4 Note that the whiskers reflecting the median 75% of the precincts are now much narrower. Whereas the range in Figure 5.1 spreads from 0 all the way to 1, the majority group ethnic vote share in a precinct rarely exceeds 75%, and the minority group vote share is rarely above 35%. This makes sense, given the demographic circumstances of each group. Both majorities and minorities have their maximum support levels capped by demographics. By definition, majority groups have larger numbers, so their cap is less restrictive. But they are rarely the only group in a given

4It also includes those parties which represent Serbs in either type of canton, but this number is small enough as to be practically insigniﬁcant.

133 Cantonal Level Cantonal Level Majority Groups Minority Groups 1.00 1.00

0.75 0.75

0.50 0.50

0.25 0.25

0.00 0.00

Vote Share for Ethnic Parties Cantonal State Vote Share for Ethnic Parties Cantonal State Assemblies House of Assemblies House of Representatives Representatives

Figure 5.2: Ethnic Voting in Bosnia by Group Type district. Even in a district where everyone voted for ethnic parties, the majority group would still not win 100% of the vote, as a small number of ethnic minority group members would vote for parties of their own group, not the majority’s. The cap is much more restrictive for minority group members. Since minority members by deﬁnition constitute less than half of their canton, it would be extremely surprising for parties representing them to exceed the 50% barrier. While they may reach relatively high numbers due to their geographic concentration within precincts, this is quite rare, and results in a relatively low median vote share by precinct. Comparing aggregated means, while suggestive, ignores the enormous variation between precincts. Figure 5.3 shows a scatterplot of ethnic voting. Those dots on the 45◦ line represent precincts where voters supported ethnic and non-ethnic parties in equal proportions in elections for Cantonal Assemblies and the House of Representatives. Dots above the line are those precincts which supported ethnic parties at a higher level in the House of Representatives than they did in the Cantonal Assemblies, and those below the line represent lower levels of ethnic party support in the House of Representatives than in the Cantonal Assemblies. While a large number of precincts are concentrated on the 45◦ line, a signiﬁcant number of precincts deviate substantially, suggesting a high degree of split-ticket voting between ethnic and non- ethnic parties. Moreover, points are scattered both below and above the 45◦ line, suggesting that some voters have switched their votes in favor of ethnic parties at the state level, while

134 Split Ticket Voting in Bosnia and Herzegovina 1.00

0.75

0.50

Ethnic Voting in 0.25 House of Representatives

0.00 0.00 0.25 0.50 0.75 1.00 Ethnic Voting in Cantonal Assemblies

Figure 5.3: Ethnic Voting Scatterplot others have moved away from ethnic parties. At the extremes, in the upper left corner and the lower right quadrant, are those precincts which saw almost total shifts between ethnic and non- ethnic parties, with virtually every single voter switching their vote from ethnic to non-ethnic (or vice-versa) in a single election. Figure 5.3 shows that even in a country like Bosnia, where ethnicity is extremely relevant to many aspects of social and political life, identity alone is not suﬃcient to predict voting behavior. Substantial numbers of voters are voting splitting their tickets between ethnic and non-ethnic parties even in the context of the same election. Of the 8,311 precincts covered in the study, 1,272 saw a change in ethnic voting (in either direction) of more than 25%. In the next section,

I test 5.2, 5.30, and 5.31 using regression analysis.

5.3.2 Local majority/minority status as independent variable

I use ecological regression analysis with ballot-level ﬁxed eﬀects in order to identify patterns in split-ticket voting to test 5.2, 5.30, and 5.31. In the absence of individual-level data, ecological regression represents the best possible way to generate estimates of individual-level ticket

135 splitting. I therefore model three separate regressions using the following model speciﬁcation:

H = G :8 : + :8 + :8

where : indexes cantons-years and 8 indexes voting precincts. The dependent variable H is the diﬀerence in ethnic voting in precinct 8 in canton : between the cantonal legislative election and the state legislative election. A value of H:8 greater than zero indicates that ethnic parties did better at the state level than they did at the national level, whereas a negative H:8 indicates that non-ethnic parties did better at the state level than the cantonal level. When H:8 = 0, support for ethnic parties and non-ethnic parties did not change between cantonal and state-level elections.

Three separate regressions are run with the independent variable G:8 representing the share of the vote going to ethnic parties representing the cantonal-level majority group, the cantonal- level minority group, or non-ethnic parties. The models must be run separately, as putting all three measures into a single model speciﬁcation introduces near-perfect colinearity. Since all votes are coded as belonging to one of the three categories, vote shares of all three measures sum to one in almost all precincts.

OLS estimates of therefore produce estimates of the percentage of the voters who voted for either ethnic majority parties, ethnic minority parties, or non-ethnic parties at the cantonal level but switched their vote at the state level. The sign on the coeﬃcient also indicates the

“direction”—in this case, whether in favor of ethnic parties (a positive ) or in favor of non- ethnic parties (a negative ). The interpretation of —i.e., that a one unit change in G generates a unit change in H— is the probability that a single voter will change their vote between cantonal levels and state levels; because G measures vote share, a one-unit change in G corresponds to going from zero voters to all of the voters. Ecological regression of this form has been used in other studies of voting behavior, where voter characteristics be relevant but are diﬃcult to measure at the individual level. It has been used most commonly in the American context to study voting diﬀerences based on gender or race in heterogeneous electoral districts.5

Indexing by : allows for intercepts to vary by canton, and controls for some important

5See Gosnell (1957), Loewen & Grofman (1989), Freedman, Klein, Sacks, Smyth & Everett (1991), Grofman & Davidson (1992), King (1997) and Cho (1998).

136 OLS Analysis Dependent Variable is Change in Ethnic Voting Between House of Representatives and Cantonal Assembly Elections Non-ethnic 0.265*** Parties (0.068) Minority -0.326** Parties (0.115) Majority -0.109* Parties (.044) Adjusted '2 0.459 0.452 0.401 = 6483 6483 6483 Standard Errors in Parentheses. *:p< .05; **:p< .01; ***:p< .001 Model includes Canton-Year fixed effects (Coefficients not shown) Standard errors are clustered by Canton-Year

Table 5.3: Split-Ticket Voting Regression Results theoretical concerns and other possible explanations. Burden & Kimball (2009) argue that split-ticket voting in the United States is a result of contest-specific factors, especially candidate quality, influencing individual voters. Given that American elections allow voters to vote for individuals, not parties, an especially popular or competent candidate may be able to overcome partisan attachment in a specific race (or equally possible, an especially incompetent or unappealing candidate may be able to drive co-partisans to the other side of the aisle). While somewhat lessened in the Bosnian context, this concern is not completely irrelevant. Bosnian political parties have very strong brands, and are generally controlled by strong and recognizable party leaders. The highly fragmented party system and relative ease of starting a new political party is such that strong individual candidates typically form their own parties if they develop a large enough following on their own. Nevertheless, an open-list system does give Bosnian voters the chance to cross party lines at different levels of government, which they may very well do if a specific candidate on a party list is highly appealing. If, for example, a particularly appealing non-ethnic candidate was campaigning at the state level, those who voted for ethnic parties at the state level but switched to vote for that particular candidate would drive increases in H correlated with G, introducing omitted variable bias. Including fixed effects in the model controls for this effect as candidate quality would be constant at the cantonal level. Since cantons are the smallest level of election in this study, canton-year dummy variables are equivalent

137 to ballot dummies. Table 5.3 shows the results of these regressions. The positive coefficient on the non-ethnic parties suggests that a voter who supports a non-ethnic party for a cantonal-level election has a roughly one in four chance of switching her vote to support an ethnic party at the state level, which supports 5.30. Since these voters are the majority group within their canton, ethnic parties remain viable contenders for political office with a high chance of winning on the basis of ethnic mobilization at both cantonal levels and state levels. But since the state is much more ethnically diverse than the canton, these groups stand to gain more through ethnic representation at the state level than they do at the canton, and thus have stronger incentives to support ethnic parties there. The negative coefficient on the Ethnic Minority Party term confirms 5.2. The estimate indicates that roughly one in three people who support minority ethnic parties at the canton level switch their vote to support non-ethnic parties at the state level. The theory predicts that those supporting canton-level ethnic majority parties should have the weakest incentives to change their vote, as any reason to support the ethnic majority party at the cantonal level is only stronger at the state level. This group of voters shows the smallest likelihood of switching their vote, at only one in ten. While all of these values are statistically significant at the .05 level, the final coefficient on canton-level majority groups only barely meats the standard (? = .05003). Increases in ethnic voting at the state level should be driven by votes going to parties that represent cantonal majority groups. Since those groups have a much greater likelihood of actually winning office, ethnic majority group members are more likely to support ethnic parties than ethnic minority group members. The dependent variable specified in the three models above is the difference in ethnic voting, and cannot differentiate between ethnic voting increases in majority or minority groups. I therefore run two more regressions, this time using difference in canton-level majority and minority ethnic party support as the dependent variable, and the non-ethnic party share at the cantonal level as the independent variable. Table 5.4 shows these results, and confirms the predictions of 5.31. Roughly one in five voters who support non-ethnic parties at the cantonal level switch their vote to support parties of the local level majority group, whereas only one in seventeen voters who supported non-ethnic parties switched to support canton-level minority parties.

138 Analysis of Increases in Ethnic Voting Change in Majority Change in Minority Ethnic Party Vote Ethnic Party Vote Share Share Non-ethnic 0.205** 0.059* Party Share (0.063) (0.030)

Adjusted '2 0.609 0.301 # 6483 6483 Standard Errors in Parentheses. *:p< .05; **:p< .01; ***:p< .001 Model includes Canton and Year fixed effects (Coefficients not shown) Standard errors are clustered by Canton-Year

Table 5.4: Disaggregated Dependent Variable

5.3.3 State level demographic status as independent variable

I have argued that local level demographic circumstances are strong predictors of ethnic voting behavior, and that majority or minority status should affect all voters regardless of which specific group they belong to. In other words, it is less important whether a voter is a Croat or a Bosniak than whether they are a member of a cantonal minority or majority group. If we could hold the voter’s ethnic identity constant, but vary whether that identity group is a majority or minority group within their district, we should observe variation in voting behavior. While it is not possible to randomize group status, it is possible to compare group behavior across regions, exploiting the fact that in the Bosnian context the same groups can be majorities or minorities depending on which canton they find themselves in. I do this in Figure 5.4. These plots show the differences in the coefficients discussed in Table 5.3 disaggregated by ethnic group. These wider confidence intervals around Croat cantons are a result of different sample sizes. Since Croats approximately 15% of the population of the country, whereas Bosni- aks are roughly 50% of the country, there are more Bosniak-majority precincts in the dataset than there are Croat-majority precincts. Disaggregating by specific ethnic group thus results in more precise estimates for Bosniaks than it does for Croats. Nevertheless, the majority/minority distinction seems to matter more than specific ethnic identification. The coefficients for majority share in Bosniak cantons and majority share in Croat cantons overlap (i.e., a Croat living in a Croat canton who voted for a Croat party at the cantonal level has roughly the same chance of

139 Change in Ethnic Voting Between Cantonal and State Parliaments

Majority Share in Bosniak Cantons

Majority Share in Croat Cantons

Minority Share in Bosniak Cantons

Minority Share in Croat Cantons

Multiethnic Share in Bosniak Cantons

Multiethnic Share in Croat Cantons

-0.50 -0.25 0.00 0.25 0.50

Figure 5.4: Comparison by Group changing her vote to a non-ethnic party as a Bosniak living in a Bosniak canton who voted for a Bosniak party). Likewise, the coefficients for minority share in Bosniak cantons and minority share in Croat cantons overlap (i.e., a Croat living in a Bosniak canton who voted for a Croat party at the cantonal level has roughly the same chance of changing her vote to a non-ethnic party as a Bosniak living in a Croat canton who voted for a Bosniak party). The differences between groups given majority or minority status are not statistically significant. However, the differences between majority and minority do not overlap, as the differences between them are statistically significant. All this suggests that a local level minority group member is more likely to behave the same way as local-level minority group member of the other ethnicity than she is as coethnics living in areas where they are the majority. When it comes to ticket-splitting, local demographics matter more than group attachment. Things are murkier when looking at the multiethnic share, as seen in the lowest two line

140 Ethnic Voting at Cantonal Level

Croat Party Support in Croat Cantons Bosniak Party Support in Density Bosniak Cantons

0.00 0.25 0.50 0.75 1.00 Vote Share Ethnic Voting at State Level

Croat Party Support in Croat Cantons Bosniak Party Support in Density Bosniak Cantons

0.00 0.25 0.50 0.75 1.00 Vote Share

Figure 5.5: Ethnic Vote Share Precinct Distribution segments of the figure. While both coefficients are positive and statistically significant after disaggregating by ethnic identification, there is no overlap between them, suggesting a statistically significant difference in the magnitude of split-ticket voting between the two groups. This would imply that that those living in Bosniak cantons are actually more likely to switch their vote to ethnic parties than those living in Croat cantons. While this is borne out by the data, this is likely the result of ceiling effects. The analysis here measures split-ticket voting to gauge the relative weight voters place on ethnic labels when voting, and as such is susceptible to ceiling and floor effects. In the Bosnian case, this is a more serious issue among Croat voters than Bosniaks.

141 The issue is that even though both Bosniaks and Croats show similar propensity for split- ticket voting, they differ in their likelihood to support ethnic parties in the first place. Figure 5.5 shows a smoothed distribution plot of vote share going to ethnic parties representing the cantonal-level majority group at both levels of elections. The top chart in the figure is the cantonal elections, while the bottom chart is the state level elections. The x-axis ranges from zero to one, (i.e., those precincts where ethnic parties received no votes on the left side to those where majority groups ethnic parties received all the votes on the right), and the y-axis represents the probability density in all the data set. The red area represents the votes going to Croat ethnic parties in Croat cantons, while the green area represents the votes going to Bosniak ethnic parties in Bosniak cantons. The two distributions skew in different directions. The Bosniak distribution is single-peaked with a skew towards 0, while the Croat distribution is saddle shaped, but skewing to the right. This suggests a difference between the two groups. This is consistent with other scholarly work on Bosnia, which frequently finds that Croats tend to be much more ethnically-oriented than Bosniaks within the Federation (Džankić 2015, Hulsey 2015, Keil & Perry 2015, Zdeb 2016). Candidates in the Croat community often rally voter support on the creation of parallel institutions and demands for the creation of a third entity, separating the Bosniak and Croat sides which currently share a federal entity between them. Ethnic parties on the Croat side are also more radicalized and face weaker non-ethnic or ethnically ambiguous competition. That the Croat side is more radically ethnified creates the potential for bias in the estimations in Croat cantons. Nevertheless, the “push” between cantonal levels and state levels moves both groups of people to the right. The large bulge on the left of the distribution at the cantonal level is noticeably more to the right at the state level, suggesting an increase in ethnic voting. That same movement in the Croat distribution is detectable, although it is pushed up against the edge of the figure by ceiling effects. A more plateaued distribution on the right side of the figure at the cantonal level becomes a peak pushing up against the 100% mark at the state level. However, since the vote share cannot exceed 100%, estimates produced for Croat voting are likely to be lower than Bosniak voting.

142 Table 5.5: Voting Returns from Odžak 2010*

Classiﬁcation Cantonal Level State Level Croatian Democratic Union (HDZ) Croat 1955 1927 Croatian Party of Rights/HDZ 1990/New Croat 1321 1493 Croat Initiative Party of Democratic Action (SDA) Bosniak 1382 972 Social Democratic Party (SDP) Non-ethnic 843 1055 Party for Bosnia and Herzegovina (SzBiH) Non-ethnic 439 294 People’s Party for Work and Betterment Non-ethnic 307 46 (NSRzB) Union for a Better Future - Fahrudin Radončić Non-ethnic 192 347 (SBB) Posavina Party Non-ethnic 158 0 Party for the People of Bosnia Non-ethnic 146 0 Total Majority Share Croat 48% 54% Total Minority Share Bosniak 20% 15% Total Non-ethnic Share Non-ethnic 32% 31% *Table shows only parties gaining 100 votes or more

5.4 The Results in a Single City: the Case of Odžak

The general pattern documented in the statistical analysis above can also be observed directly in individual cities. Consider the city of Odžak, a small, predominantly Croat municipality in the Croat canton of Posavina in the north-east of the country. Odžak is near the three-part border between Croatia, Serbia, and Bosnia, less than ten kilometers away from the Croat border, and less than one hundred kilometers away from the Serb border.6 Odžak is typical of many cities in the Federation: it has a clearly dominant demographic group (Croats, in this case), but large numbers of ethnic minorities (predominately Bosniaks) that are politically signiﬁcant. The party system is heavily fragmented, with large numbers of parties contesting elections at each electoral contest. Ethnic parties claiming to represent both ethnic groups and non-ethnic parties campaigning on a wide range of non-ethnic social issues are all present on the local ballot. Table 5.5 shows the vote tallies for the 2010 canton- and state-level elections in Odžak. The Croat ethnic vote, by far the biggest electoral bloc in the city is divided between two parties: the HDZ, and a coalition of HDZ break-away parties. The Bosniak ethnic vote is mostly con-

6Odžak is a municipality, but the unit of analysis in the regression is the precinct. In the regression, Odžak is represented by the 22 voting precincts of the city. The example uses the municipality level as a more intuitive way of understanding how strategic voting impacts outcomes.

143 solidated into a single party—the SDA. The non-ethnic side is highly fragmented. Three of the parties would meet the standards of “ethnically ambiguous” outlined in the previous chapter: The Party for Bosnia and Herzegovina, and the Union for a Better Future are both strongly associated with Bosniak candidates, and tend to do well in Bosniak areas. The People’s Party for Work and Betterment is an ostensibly non-ethnic party that campaigns almost exclusively in Croat areas. Three parties are truly non-ethnic: the Social Democratic Party is an anti-nationalist successor party to the Yugoslav communists, the Posavina Party is a multiethnic party advocating for regional interests, and the Party for the People of Bosnia is a minor populist multiethnic party originally founded in nearby Zenica that had expanded to contest elections in Posavina. For most parties there is extremely little consistency in the vote totals between the canton and state levels. The HDZ is the only party to swing less than 10% between the two levels of elections, and some parties sway by hundreds of votes—a high level of volatility in a city of well under 10,000 voters. But the gains and losses between the two levels of elections are not random. In total, the Croat ethnic vote goes from 48% to 54%. Since Croats are the majority group in the canton but a minority in the state, it stands to reason that they have much more to gain through ethnic representation at the state level, and will therefore be more likely to support ethnic parties at that level. The Bosniaks, on the other hand, are the minority group in the canton, and thus have more to gain at the cantonal level than the Croats, and face smaller barriers to entry at the cantonal level. This helps explain why the Bosniak share of the vote is higher at the cantonal level than the state level—the opposite pattern from the Croat ethnic parties. The non-ethnic party’s vote share remains relatively constant, but it is clear that this is the result of Croat ethnic ticket-splitting and Bosniak ethnic ticket-splitting canceling each other out. The SDP seems to have gained many votes while the SDA has lost many, which would be consistent with Bosniaks supporting an ethnic champion at the cantonal level, but multi-ethnic socialists at the state level as a second-best option. This may also explain the increased vote tallies of the SBB, who campaigns on an ostensibly multi-ethnic platform but still targets mostly Bosniaks.7

7The SBB is one of the few parties in the dataset that changes classiﬁcations between ethnic and non-ethnic. In 2010, the party was ostensibly non-ethnic. See next section.

144 Figure 5.6: Bosnian SNSD Logo, adopted 1996

5.5 Variation over Time

The results presented above suggest that for political parties making strategic decisions about how to campaign to win elections, the optimal strategies depend on both constituency and the level of government for which a party is campaigning. For the State House of Representatives, the optimal strategy for winning is to mobilize an ethnic majority group within a canton using ethnic appeals. If a party is trying to win a cantonal assembly, then a non-ethnic appeal targeting a multi-ethnic voting base is a more viable strategy, as doing so will be likely to win votes from both cantonal majority and minority groups. The ecological regression research design employed above is ill-equipped to observe and test this tendency, as it cannot detect trends over time. In the time period under study, only one party ever makes demonstrative efforts to change its appeals from ethnic to non-ethnic. In Chapter 6, I turn to the case of Latvia to document that elites actually do adopt new strategies regarding ethnic appeals in light of the incentives I document here. Nevertheless, there is some qualitative evidence of the parties that have substantially changed their ethnic appeals that support this prediction. In this section, I survey three parties that have made noticeable changes to their ethnic appeal strategy: the Serb-dominated SNSD, led by Milorad Dodik, the Union for a Better Future led by Fahrudin Radončić, and the Bosnian Patriotic Party, led by Sefer Halilović. The highest profile example of this change is the Alliance of Independent Social Democrats (SNSD), led by Milorad Dodik. In the early days of post-war Bosnia, the SNSD was identified by foreign development agencies and peacebuilders as a potential moderate party in the Republika Srpska. Dodik, who advocated a more moderate political platform than Karadžić’s SDS, was seen as the natural partner for the international community seeking to moderate and diffuse ethnic tensions. This initial moderation is still represented in the party’s name and logo: the name simultaneously suggests a conciliatory approach to politics (“Alliance” and “Inde- pendent,” rather than any ethnic labels), and an ideological orientation (it is a party of “Social

145 Democrats”). Its logo is also surprisingly devoid of any nationalist imagery: the stylized flower suggests an homage to the same red rose pattern used by the SDP. Despite close involvement with the international community, Dodik felt the SNSD was poised to gain votes from the SDS by attacking it as corrupt and inefficient. But in its campaigns, the SNSD combined a left-of center ideology and appeals to valence issues like incorruptibility with more extreme nationalist positions, attacking the legitimacy of the Bosnian state and the oppression of Serbs at the hands of the other two groups (Toal 2013). The ethnic rhetoric proved a highly successful electoral tac- tic, and increased in every electoral cycle until the party was warned and eventually suspended from the Socialist International for nationalist and extremist political positions that were inconsistent with the tenets of social democracy (Dnevni avaz 2011). A far cry from his initial status as the moderate partner of international peacebuilders, in January of 2017 the United States Treasury Department placed sanctions on Dodik’s personal assets in the United States, arguing that he was actively trying to subvert the peace deal in his country and attempting to create an independent Serb territory by partitioning the Bosnian state (Radio Liberty 2017). Today, Dodik is the Serb president of Bosnia, and has effectively broken all ties with Washington and Brussels, campaigning on a platform of secessionism, ethnic Serb nationalism, denial of Serb atrocities during the civil war, and closer ties with Serbia and Russia (Surk 2018, Pandora’s Box 2018, Hall of Infamy: Milorad Dodik 2019). The SNSD thus began as a moderate, socialist party (similar in ideology to the Social Democratic Party), that campaigned as a non-ethnic alternative to the ethnic Serb parties in the Republika Srpska. It became a hardline extremist ethnic party, and in doing so became the most important party representing the Serb entity of Bosnia. The electoral dynamics outlines here suggest that such a transition should not be surprising: the strategy which makes it easiest to win state-level votes in Bosnia is one of mobilizing local-level ethnic majority groups. A similar progression is seen on the Bosniak side, with the candidacy of Fahrudin Radončić and his Union for a Better Future. As mentioned in Chapter 4, Radončić is a wealthy media and construction magnate based in Sarajevo. Radončić established his own party, the Union for a Better Future (SBB) in 2009. Having run for the Bosniak seat of the presidency multiple times, Radončić seems to have more actively embraced an ethnic identity as his political career

146 The text in the middle reads, “Serb Power!” The text boxes in the graphic wall behind read “For Serb Victory” and “A Serb Future”

Figure 5.7: Bosnian SNSD Homepage, February 2017 evolves. The SBB was established as a secular, multinational party, advocating for the rights of all citizens of the Bosnian state–a position still reflected in its official statute in 2014 (SBB). Yet a stark contrast in the party’s strategy was visible on the streets of Bosnian cities between the 2010 election and the 2014 election, when Radončić changed his branding strategy in his campaign for the Bosniak seat of the Bosnian Presidency. The contrast is displayed in Figure 5.8. In 2010, Radončić described himself as “a successful person, for a successful Bosnia.” In 2014, Radončić was portrayed as a “successful Bosniak for a strong Bosnia” (emphasis added). To emphasize that voters were voting for a Bosniak, the diacritic mark above the ‘s’ in “Bošnjak” was replaced with a check mark used on electoral ballots. The 2014 election also saw the introduction of a party-wide slogan, “with the people,” used in campaign materials for presidential, legislative, and local races. While it is hard to say exactly what affect this had on Radončić’s electoral success—he lost the presidency in both elections, and actually did slightly worse in 2014 than in 20108 But it is telling that Radončić’s campaign seems to have made a conscious decision to portray him using ethnic labels as he sought state-level office. In this way, Radončić exhibits

8This may have been attributable to the strong showing of the DF in the election, which had not competed in the 2010 election, but came in third in 2014.

147 2010 2014 “A successful person “A successful for a successful Bosnia!” Bosniak for a strong Bosnia With the People”

Figure 5.8: Campaign Posters for Bosnian SBB a less extreme version of the trajectory followed by Dodik and the SNSD: increasing ethnic appeals in order to campaign at the state parliament and presidential levels. Both Dodik and Radončić show a tendency for increased ethnic appeals as they rise to prominence at the state level. But there is also at least one example of the opposite tendency: for candidates to employ fewer ethnic appeals in order to contest elections at the local level only. Sefer Halilović, also discussed in Chapter 4, was a Bosniak political and military leader, often seen as an opposition leader to the SDA within the Bosniak community. His military campaigns earned him name recognition and prestige in the Bosniak community, as well as an indictment for war crimes from the International Criminal Tribunal for the Former Yugoslavia (he was eventually acquitted of all charges). The founder of the Bosnian-Herzegovinian Patriotic Party-Sefer Halilović (BPS), Halilović has led the party through several elections, running for president

148 2006 2014 a b

Figure 5.9: Logos for Bosnian BPS himself in 2014 and earning the party’s highest ever share of the vote: 8%. Despite never winning the presidency, the party has had some success in the state parliament, and local-level governments. But over time, as it has become clearer that the party’s successes will be mostly at the local level rather than the state level, the party has distanced itself from its explicitly Bosniak roots. Originally founded by a Bosniak military leader famous for being accused of atrocities against Croats (International Criminal Tribunal for the former Yugoslavia 2005), the party has undergone a substantial “rebrand” seen in Figure 5.9. The original logo in Figure 5.9 shows a prominent Bosniak fleur-de-lys. The new logo in Figure 5.9b, implemented before the 2014 election, removes nationalist identifiers in favor of a generic light bulb icon, suggesting innovation rather than ethnic representation. The rebrand is also present in the new party program, which defines its origins not as a Bosniak organization, but as an anti-fascist political movement, which sought only to defend Bosnian statehood–an arrangement it claims benefits all groups (BPS - Sefer Halilović 2016). Parties changing from ethnic to non-ethnic or vice versa are relatively rare in Bosnia. Most ethnic parties have been that way since their founding, and have not substantially changed in their explicit advocacy of ethnic group interests, or in their self-identification as members of a specific ethnic group. These examples suggest that when parties do adopt ethnic identifiers that they previously did not employ, it is when they are seeking national-level office. Both observed cases of politicians who changed strategies towards more explicit ethnic identification

149 and more extreme ethnic rhetoric did so in pursuit of office in the state-level parliament and the presidency, the highest offices in Bosnia. Abandoning ethnic identifiers happens when parties are declining in state-level relevance, and instead concentrate on winning lower-level races like local and cantonal assemblies.

5.6 Conclusion

Bosnia is one of the most ethnically polarized countries in the world. Large sections of daily life are segregated by ethnicity, and reminders of ethnic tensions are ever-present. Politicians routinely stir up ethnic grievances, and for many people in the country extreme and highly traumatizing experiences with ethnic violence are real and present memories. Yet even in in this highly contentious environment, where ethnic identities are a fact of life, voters prove surprisingly flexible in their willingness to support ethnic parties. Non-ethnic parties can and do win elections in Bosnia, and a large number of voters are responsive to claims that ethnic identities should not be the basis of political action. Ethnic identities are not the only possible basis of political activity in Bosnia. It is not the case that Bosnian voters view themselves in either ethnic terms or non-ethnic terms at all times. Rather, Bosnian voters have both ethnic and non-ethnic identities which overlap and complement each other, and can be used as the basis of political mobilization depending on circumstances. When ethnic representation stands to benefit them the most, and when ethnic parties are the most likely to win real political power, Bosnian voters are more likely to support parties that claim to champion ethnic interests. If they have little to gain through ethnic representation, either because the benefits they would receive are slight, or because their group has no chance of gaining access to political power, they will respond to non-ethnic appeals. Many voters can even vote for two candidates espousing radically contradictory ideologies on the role of ethnicity in politics even in the context of the same election. All of this suggests that identity is not deterministic on voting behavior, but instead complements the way voters perceive their circumstances, interests, and options. Eth- nicity in Bosnia is clearly important, but ethnic representation is a means to an end, not a goal in and of itself.

150 This chapter has examined the use of ethnicity and the choices voters make when presented with plausible ethnic and non-ethnic options for political representation. In the next chapter, I turn my attention to the choices made by elites, and whether they choose to mobilize voters on ethnic grounds or not.

5.7 Appendix: Robustness to Alternate Coding: Voter Per-

ceptions over Candidate Declarations

Perhaps the biggest concern with this analysis is that it relies on a coding scheme which privileges the ways that parties describe themselves. However, since the theory rests on the strategic calculations of voters, the ideal measurement would be the way that voters perceive the parties, not the way that parties try to describe themselves. This is a particular concern, since the Bosnian media tends to describe several of the parties I have coded as non-ethnic as if they were supported exclusively by single ethnic blocs. I therefore test the results against alternate coding schemes that take this into account. For some parties, there is absolutely no reason to question whether or not they are ethnic. Parties like the Bosniak SDA, and the Croat HDZ are nearly universally accepted to be explicit ethnic group advocacy parties, and make that advocacy central to their party programs and electoral campaigns. There are also some parties which are undeniably non-ethnic. The Social Demo- cratic Party and the Democratic Front openly broadcast that they are running multiethnic candidate lists and have declared themselves as left-leaning alternatives to Bosnia’s ethnic blocs. Some of these parties go so far as to run multiple candidates for the various ethnic seats of the presidency in order to make it clear that they are privileging no single ethnic group over others. The concern, then, is in the parties which describe themselves as non-ethnic, civic-minded parties but in fact draw nearly all of their support from a single ethnic group. If voters perceive these to be ethnic parties despite the parties’ own rhetoric, then the strategic calculation theory could be called in to question. I therefore created an alternate coding scheme that is more generous in assigning ethnic aﬃliation. On the Bosniak side, the biggest changes to the coding scheme concern Haris Silajdžić’s Party for Bosnia and Herzegovina (SzBiH) and Sefer

151 Regression Results–Alternate Coding Scheme Dependent Variable is Change in Ethnic Voting Between State and Canton-Level Non-ethnic 0.125*** Parties (0.030) Minority -0.126* Parties (0.054) Majority -0.075* Parties (0.029) Adjusted '2 0.289 0.275 0.261 # 6483 6483 6483 Standard Errors in Parentheses. *:p< .05; **:p< .01; ***:p< .001 Model includes Canton and Year fixed effects (Coefficients not shown) Standard errors are clustered by Canton-Year

Table 5.6: Robustness to Alternative Codings

Halilović’s Bosnian Patriotic Party. Both parties are closely tied to the personalities of their founders, both of whom were actively identified with Bosniak military forces and have been extremely vocal in their condemnation of Serb war criminals against Bosniak civilians. While their parties generally advocate for European principles of civic national identity for Bosnia, it is not unreasonable to suspect that at least some part of the Bosnian population may view them as Bosniak nationalists. On the Croat side, the biggest change is the People’s Party for Work and Betterment (NSRzB), founded by Mladen Ivanković-Lijanović. While more sedate in his explicit identification as an ethnic Croat, Ivanković-Lijanović has been the NSRzB’s only presidential candidate, contesting only the Croat seat of the Bosnian Presidency, and media sources often describe the NSRzB as a party whose support derives predominately from Croats. Having modified the party codings, I run the regressions again on the change in ethnic party vote share. The results are listed in Table 5.6. Comparing with the results shown in Table 5.3, it is clear that the alternative coding scheme does have an effect on the coefficient estimates. The estimates are greatly diminished, taking the value of roughly half of the estimated values in the standard coding. However, the general pattern predicted by the theory remains, albeit in diminished form. The coefficient on the majority party variable is much smaller than the minority or non-ethnic party variables, and the signs on the coefficients are in the expected directions. Essentially, what this means is that the strategic vote-switching seen in Bosnia, even

152 when limited only to the explicitly avowed multiethnic parties and regional non-ethnic parties, is still of a magnitude great enough to support the predictions of the theory. Moreover, under this alternative classification, the values on the Adjusted '2 drops substantially. This suggests that the previous coding scheme, relying on the explicit declarations of candidates and parties, explains a substantially larger proportion of the observed variation in Bosnia ticket splitting than this alternative coding that relies on subtext and assumed perceptions. The coefficient estimates are smaller because those voters switching from the SzBiH—an avowedly non-ethnic party with a solid mono-ethnic support base—to the overtly Bosniak SDA under this coding is ignored as unimportant. This coding would, however, produce a larger coefficient on the multiethnic support base if large numbers of voters were switching from avowed multiethnic parties like the Social Democrats to the SzBiH, which the estimates suggest is not happening. These findings shed some light on a pervasive puzzle in Bosnian politics. If parties like the SzBiH rely on a single ethnic base, why do they insist on pretending to a non-ethnic identity justified by ethnicity-blind rhetoric? These results suggest that the advocacy of civic over ethnic nationalism may be an asset in attracting voters who are strategically concerned about wasting their votes on non-viable ethnic parties.

153 6

Ethnic Politics in Latvia

In this chapter, I turn to the case of Latvia, and show how the voter-level phenomena which I documented in the previous chapters impact the long-term development of a party system in a country consolidating democratic institutions. Bosnia is a consociational democracy with overlapping levels of governance. Ethnic diversity is managed by institutions consciously engineered to facilitate the representation of all relevant ethnic groups in a multi-ethnic democracy. In contrast, Latvia is a straightforward parliamentary system. It is a small, highly centralized state using proportional representation reminiscent of other European parliamentary democracies like Austria, Denmark, and the Netherlands. Whereas Bosnian institutions are intended to ensure equal representation among all three of the country’s constituent peoples, the constitution of Latvia grants no special consideration to the interest of ethnic minorities. In fact, the bulk of Latvia’s constitution institutional structure predate the ethnic diversity of the country as it exists today. Whereas Bosnia’s transition from socialism was accompanied by a devastatingly destructive civil war, Latvia is often seen as a “success story” of post-communist transition. It declared independence from the Soviet Union in 1991, and successfully installed a democratic regime in the early 1990’s. After meeting European standards of democracy, market liberalism, and human rights, it joined the European Union in 2004, and adopted the Euro in 2014. Despite having been the site of several important Soviet military and naval installations as recently as the early 1990’s, Latvia joined NATO in 2004. This transition from Soviet republic to European

154 democracy has been accomplished with virtually no serious inter-ethnic violence or military confrontation. The peaceful nature of the transition is all the more remarkable given the tension caused by Latvia’s enormous Russian-speaking minority. Largely the result of Soviet-era migration, ethnic Latvians today constitute only a bare majority of the country’s population. Today, only three-fifths of Latvia is ethnic Latvian, with the remainder mostly Russian-speaking Slavs. In this chapter I examine the ethnic diversity of Latvia, with the goal of identifying how ethnic divides are politicize by elites seeking to mobilize voters and win office through democratic elections. In the first section I briefly overview the historical background of the ethnic divide in Latvia, drawing attention to the divergent languages and historical origins of the two groups. The legacy of colonialism and the domination of ethnic Latvians by outsiders is especially important in post-Soviet campaigning, as many political entrepreneurs use this narrative to draw distinctions between ethnic groups and argue for the necessity of group representation. I then discuss the transition from communism to democracy at the end of the twentieth century, and how that transition established many of the relevant political actors and institutions which structure ethnic conflict and constrain the choices of Latvian political entrepreneurs today. I then examine Latvian political campaign strategy in the post-Soviet period, showing how political parties adopt names, positions, and rhetoric which either embraces a specific ethnic identity, rejects ethnicity as a relevant political distinction, or ignores the issue entirely. The final section concludes.

6.1 Historical Background

Latvia is generally understood to be divided between two groups: “Latvians” and “Russians.” Yet the designation “Russian” is not entirely accurate, since many “Russians” are actually Russian-speakers who may or may not consider themselves ethnically Russian. According to the most recent Latvian census, ethnic Latvians comprise 62.07% of the population. Ethnic Russians are by far the largest ethnic minority group, at 26.91% of the population, followed by Belorussians (3.29%), Ukrainians (2.21%), Poles (2.16%), and Lithuanians (1.18%) (Central Statistical Bureau of Latvia 2011). Unlike in Bosnia, the ethnic divide in Latvia is a product of

155 relatively recent demographic change. While the area known as Latvia has long been a meeting point between ethnic Germans, Balts, and Slavs, the bulk of the ethnic minority population today are those who arrived during the period when Latvia was a part of the Soviet Union and their descendants. The majority group and minority group share no recent common ethnic ancestor, and differ not only along ethnic cleavages but also linguistic and religious lines. Russian and Latvian are written in different alphabets, come from distinct linguistic families, and are in no way mutually intelligible. Among ethnic Latvians, Lutheranism and Catholicism are the most common religious traditions, while the Latvian Orthodox Church is the most common religion among ethnic Russians. Some Latvian historians claim that the defining feature of Latvian ethnic and national iden-

tity is the history of subjugation by foreign powers (G, ermanis¯ 2007, Kalnins 2015). While the Latvian people and their ethno-linguistic ancestors have lived in the region around the Baltic Sea for millennia, a sovereign Latvian political authority has existed for only a few decades throughout all of recorded history. Latvian political entrepreneurs frequently draw attention to the rarity of ethnic Latvian self-determination as a reason justifying ethnic representation when campaigning for votes. Four separate historical periods are crucial for understanding the history shared by ethnic Latvian ethnic identity: the pre- and early-modern domination of the Baltic territory by German aristocrats and the Russian Empire, the brief period of independent Lat- vian statehood following World War I, the communist period under which Latvia was forcibly incorporated into the Soviet Union, and the transition to democracy in the 1990’s. People speaking Baltic languages that today are recognized as dialects of Latvian and Lithua- nian can be placed in the area around the Baltic Sea at end of the last ice age. While Latvian and Lithuanian today share a high degree of overlap and mutual intelligibility, they are completely distinct from Estonian (a Finno-Ugric language closer to Finnish and Hungarian) and the Slavic and Germanic languages. One of the last pockets of Europe to Christianize, the Baltic people were ﬁrst converted to Catholicism following a crusade in the thirteenth century by the Teu- tonic Order. These German knights subjugated the Baltic tribes and installed themselves as feudal lords over the conquered territory. These Livonian territories comprised independent duchies, similar to the other German feudal divisions until the late eighteenth century, when

156 the territory was annexed by the Russian empire under Catherine the Great. The feudal system imposed by the German knights was essentially serfdom reinforced by ethnic and linguistic differences: at the top where a German-speaking warrior class and their aristocratic descendants, while the lower classes were Baltic-speaking peasants. Latvian ethnic and national identity emerged under the limited liberalizing reforms of the Russian empire in the 19th century. Alexander I placed limits on feudal rights exercised by the German nobility in the early 19th century, which freed many rural laborers to migrate away from the agricultural fields. As these ethnic Latvians migrated to cities, they not only entered into bourgeois professions, but also began studying at Baltic universities. While cities had been almost exclusively German and Russian speaking, these waves of migrants brought a substantial Baltic-speaking population into cities like Riga. When czar Alexander II eased restrictions on political organizing and publishing, many of these Latvian-speakers began publishing Latvian- language newspapers, pamphlets and volumes. This era is often referred to as the “Awakening,” and is the first time a distinct Latvian ethnic identity emerges as a cultural and social force. It is during this time that the peasants in Latvia began to be known as latvieši (“Latvians”), when they had previously been known simply as undeutsche (“not-Germans”). This also dis- tinguished the Latvian people from the other Baltic groups. The boundary-making process of these identity entrepreneurs also absorbed three separate regional identities into a single ethnic and national identity. Plakans (1974) writes that the word latvieši and its eventual entrance into common parlance “was meant to suggest the existence of a Latvian tauta (Volk),” uniting the inhabitants of Livland, Kurland, and Vitebsk into a single identity under a shared culture and common historical descent. Intellectuals and scholars in this period codified a standard “Lat- vian” language, relegating Livonian and Curonian to status as dialects of an official high Latvian language which they argued had the potential for literary and scientific achievement on par with Russian and German. Zake (2007) argues that this identity formation did not initially have any substantial political ideology or agenda, but was rather the result mainly of ethnic Latvian intellectuals and students attempting to reconcile their own perceptions of cultural and ethnic distinction from ethnic Germans and Russians following the process of modernization and urbanization. Given

157 the overwhelming dominance of Germans in the universities and cultural institutions of Latvia in the 19th century, Latvians absorbed the influence of German nationalism and adopted this intellectual approach in describing and codifying their own identity. German nationalist philosophy at the time had stressed romantic ideas of common heritage, common culture, and common language. Latvian identity emerged as a result of applying this template to existing Latvian circumstances: if the Germans had a common way of life, then the Latvians must have one, too. Since the non-German people of Latvians were overwhelmingly agrarian, Latvian identity was understood to have a fundamentally rural character, and Latvian nationalist intellectuals put a heavy emphasis on documenting supposedly Latvian folk songs, stories, and pre-Christian Lat- vian mythology. The common history of the Latvian people was described in romantic terms in relation to what differentiated them from the land-owning and urban German upper classes. Ijabs (2014) argues that Latvian ethnic and national identity should be understood less as a political movement, and more as an example of post-colonial mimicry as described by Bhabha (1994). Here identity is formed by adopting the presumably universal aspects of the culture of the colonizer (the Germans), while also internalizing and accepting the fundamental “otherness” of the colonized (the Latvians). The earliest expressions of Latvian identity were not strongly linked to demands for political autonomy. Early ethnic Latvian intellectuals were also often extremely supportive of Russian administration, despite today’s Latvian nationalists’ often pro-western and anti-Russian positions. The Young Latvians, the intellectual movement based in Tartu most strongly associated with this first wave of Latvian identity formation, often expressed sympathy for Russian rule, and maintained close ties to Slavophiles and Russian nationalists. Krišjanis¯ Valdemars,¯ the most prominent of early Latvian intellectuals, published many of his tracts in Russian nationalist newspapers and journals, and spent much of his life writing in St. Petersburg and Moscow. Valdemars¯ seems to have seen Russian nationalism as entirely compatible with Latvian identity, with its focus on “non-western” ideology (i.e., distinct from German nationalism), and its emphasis on rural agrarian traditions and resistance to modernization (Zake 2007, p. 323). Moreover, the reforms of liberally-minded leaders in St. Petersburg had been the main force limiting the seigneurial authority of the German elites, and the Russian annexation of the Baltic

158 territories had facilitated the more liberal environment which allowed Latvian intellectuals to develop their identity. Despite the existence of a strong sense of ethnic and national identity going at the start of the twentieth century, the politicization of that identity and systematic movements for an independent Latvian were rare before the First World War. Even the most radical ethnic Latvian politicians usually accepted that small ethnonational groups would inevitably be absorbed by larger empires, and advocated only for a limited form of self-rule within a more democratic Russian state (Purs 2012, Chapter 1). World War I drastically changed this, with Latvians nationalists fighting a war of independence against German attempts to absorb Latvia into a unified German state, Bolsheviks attempting to expand communist control into the Baltic region, and White Russian soldiers fighting to maintain the integrity of the Russian empire. In 1920, an independent Latvian political authority signed a peace treaty with the newly-formed Soviet Union, establishing legal recognition of an independent Latvian state. The legacy of the interwar Latvian state is hotly debated by historians. The constitution enacted in 1922 established a democratic regime which extended civil liberties to Latvian citizens and insisted on popular sovereignty as the basis of government. It established a unicameral parliament, the Saiema (Latvian for “gathering”), and a relatively weak president appointed by the parliament. The country was divided into five administrative regions—the four cultural and historical regions corresponding roughly to German ducal boundaries plus the capital—each of which would elect deputies to Parliament using proportional representation. The electoral system notably created no serious obstacles for would-be officials to enter electoral contests or form their own parties. There was no electoral threshold, and a political party could legally be established by any group of five persons or more. These rules produced an enormous number of political parties. Between twenty and thirty parties were elected to parliament in each electoral cycle, requiring very fragmented coalition governments. The resulting governments were very fragile, resulting in sixteen separate cabinets in fourteen years. The system did, however, facilitate the representation of many non-Latvian ethnic minorities, and the 1922 parliamentary election saw deputies seated from parties representing Germans, Russians, Jews, Lithuanians,

159 Poles, and Old Believers.1 The interwar Latvian government eventually descended to an authoritarian regime. With echoes of Weimar Germany, economic crisis increased social tension throughout the 1930’s, including the rise of an anti-Semitic nationalist paramilitary group, the Perkonkrusts,¯ resem- bling the German Brown Shirts. In this environment, a state of emergency was declared and Prime Minister Karlis¯ Ulmanis was given emergency powers in 1936 as “leader of the people” (Hope 1999).2 Ulmanis, remembered by some historians as Latvia’s only domestically- produced dictator, is sometimes said to have ruled a “dictatorship by consensus” (Lieven 1994, p. 70). The Ulmanis regime imposed a ban on many political activities in the name of civil order, and relied on a nationalist paramilitary force, but also coopted rather than imprisoned potential political rivals. For the left, Social Democrats were given cabinet positions in the Ulmanis government, while at the same time the regime imposed laws severely restraining the organizing capacity of trade unions. For the right, the regime adopted an official stance of “Latvia for the Latvians,” and made life increasingly difficult for Germans, Russians, and Jews. Histori- ans debate whether this regime was ultimately on the path to a totalitarian regime accompanied with severe ethnic persecution (Purs 2012, p. 47). The Latvian state ceased to operate as an independent sovereign body when Ulmanis was forced by a Soviet ultimatum to sign a “Mutual Assistance” treaty with Stalin following the Molotov-Ribbentrop Pact of 1939. Within a year, the USSR had completely occupied Latvian territory and installed a puppet parliament which voted to join the Soviet Union. The Soviet period was a time of massive social, economic, and demographic change. The Latvian Soviet Socialist Republic was identified by Soviet economic planners as a strategically vital, resource-rich, but ultimately underdeveloped region of the USSR. It contained fertile lands and rich forests, and held an important port position on the Baltic Sea for both military mobilization and international trade. But it remained largely agrarian, rural, and under-educated compared to other parts of Europe. A large segment of the population also resisted Soviet incorporation for nationalist and ideological reasons (Shtromas 1999). The Soviet development

1“Old Believers” are the descendants of those who ﬂed persecution in Russia after refusing to acknowledge the Eastern Orthodox Church reforms of 1652-1666. 2Tautas Vadonia in Latvian, but often translated into German as Führer.

160 strategy had three parts: purges designed to remove potential reactionary elements from Latvian society, migration from other parts of the Soviet Union to both provide skilled labor and change the ethnic make-up of the republic, and a Russification project which would see Russian replace Latvian and German as the official language. Throughout the post-war period, large numbers of migrants came to Latvia, especially from Russia and Ukraine, concentrating mostly in major urban areas, and Latgale, the easternmost region of the country closest to Russia. As Russian language proficiency was required for employment in heavy industry and the civil service, ethnic Latvians had strong incentives to learn Russian, but few ethnic Russians learned Latvian to any high degree of proficiency (Prazauskas 1991, Misiunas & Taagepera 1993, Lieven 1994). Ethnic segregation was further compounded by Soviet centralized housing policy. Under the command economy, many of those who worked for Soviet ministries and the Red Army were housed in apartments built by those agencies directly; since Russians disproportionately comprised employees for those agencies, they also tended to live together in ethnic enclaves within urban areas (Kolstø 1995, p. 111). Even at the end of the Soviet period, the segregation between the ethnic groups was obvious: the leadership of industrial sectors was overwhelming ethnically Russian, while the agricultural sector leadership was almost exclusively ethnically Latvian (Lieven 1994, p. 291).

6.2 From Soviet Republic to European Democracy

The end of Soviet control in the Baltic states began with the economic and political reforms of the 1980’s. As part of the glasnost policy, Gorbachev’s moderately liberalizing reforms allowed Soviet citizens to criticize and question state policies in hopes of increasing eﬃciency and spur innovation. In Latvia, this new liberalized political space was quickly exploited by nationalists and independence advocates who began staging the so-called “calendar demonstrations” (Clemens 1991, Chapter 6). Commemorations of important historical milestones of Latvian statehood, or the Soviet occupation of had previously been banned by Soviet censorship, and for the ﬁrst time public events were allowed to draw attention to the historical circumstances surrounding Latvia’s incorporation into the USSR. Latvian reformers not only tried to

161 advance economic reforms, but also sought to address historical grievances over past injustices, specifically surrounding the Stalinist purges of Latvian nationalists. Much of the opposition to Soviet rule came from this historical perspective. In 1988, the Latvian Writers’ Union—the state-controlled organization which regulated creative and literary professionals—publicly de- fied long-standing state positions by declaring that there had been no popular uprising in support of Soviet annexation, and that the Soviet take-over of Latvia had been orchestrated completely from Moscow. Historians and other academics followed suit, as revisions to the official Soviet narrative spread throughout universities and publishing houses. One of the most famous political protests at this time was the so-called “Baltic Chain,” when an estimated two million people joined hands to form a human chain connecting Tallinn, Riga, and Vilnius. Protesters united the three Baltic capitals as part of a commemoration of the fiftieth anniversary Molotov-Ribbentrop Pact between Germany and the Soviet Union that ceded the Baltic states to the Soviet sphere of influence. In this way, debates over the historical record became one of the main mobilizing points for opponents of the Soviet regime. While glasnost may have been intended to encourage Soviet citizens to look for solutions for the future, many Latvians looked back to the past. Another crucial mobilizing force against Soviet Rule was in pre-modern Baltic folk traditions, and the organizations established to preserve them. Consistent with Soviet nationality policy, folk songs, dances, and national costumes were to be celebrated as a way to highlight the multicultural diversity of global communism. Concerts of traditional Livonian, Lat- vian and Latgalian songs were permitted under this policy throughout the Soviet period, and even formed part of state visits and dignitary’s processions. These performances, though, were also occasionally censored due to Soviet fears that traditional cultural expressions may also be politically sensitive. Given the strong link between these folk practices and cultural expressions and the origins of Latvian identity, there was concern that these could be used to mobilize Latvians in support of a Latvian national identity that saw itself as incompatible with, or even superior to, the Soviet class-based identity which was to triumph over bourgeois ethnic association (Dreifelds 1996, Ginkel 2002). This embrace of diversity ironically preserved ethnic Latvian culture in a way that was utilized by anti-Soviet activists following perestroika (Jubulis 2007, Agarin 2010). Members of these preservation societies participated in many of

162 the historical commemoration events, singing and dancing to traditional (i.e., pre-Soviet) songs. These non-violent, but nonetheless politically provocative activities came to define the Latvian independence movement as “the singing revolution” (Egl¯ıtis 1993). That the resistance to Soviet rule should focus so strongly on nationalist historical grievances and the violation of Latvian national sovereignty has led some to see the revival of ethno- nationalist attitudes in Latvia as a “Rip Van Winkle” story: ethnic identities had “gone to sleep” suppressed by Soviet censorship, and had now “woken up.”3 But this interpretation underesti- mates both the degree to which other, non-ethnic grievances were important points of protest and mobilization, as well as the extent to which ethnic Russians and Russian-speakers were actively involved in the protests against Soviet occupation. Specifically, environmental issues were also one of the most crucial rallying point for anti-Soviet agitators, and comprised a large multi-ethnic force opposed to the Soviet regime (Galbreath & Auers 2010). In 1986, a Moscow- proposed plan to build an enormous hydroelectric dam on the Daugava River—called ‘the river of our destiny, the river of our souls’ by a traditional Latvian saying (Kalnins 2015, p. 177)— attracted an enormous amount of scrutiny. Local activists were concerned that official party assessments had drastically overestimated the economic and technological benefits of the project by inflating projections of the electric output while underestimating its costs, specifically in the economic impact of the destroyed forest and farmland and personal displacement due to flooding. The project lead first to a massive letter-writing, public protest, and petitioning campaign which ultimately collected 30,000 signatures. The activists and experts who campaigned against the dam constituted a multiethnic organization: the Daugava River flows through several major cities in Latvia, including the capital Riga and the second largest city, Daugavpils, and has historically constituted the boundary between Latgale (with a high concentration of ethnic Russians), and Vidzeme (considered by Latvian nationalists to be the “purest” Latvian region). The project thus attracted attention from both ethnic Latvian and Russian activists (Muižnieks 1987, Rich 1987). The hydroelectric project was ultimately halted by the protests. In this way, Latvia resembled other eastern bloc countries where environmentalism was an important mobilizing force (Albrecht 1987, Jancar-Webster 1993, Fagin 1994, Steger 2004). The

3See summary in Holmes (1996).

163 Soviet economic system had long been guided by centralized planned economics, seldom valuing natural conservation or realistically assessing environmental impact. The Baltic republics especially had been enormously impacted by environmental degradation. Enormous Soviet investment and immigration had caused rapid industrialization and urbanization throughout the post-war period, yet Soviet planners had spent little eﬀort assessing the possible impact of such industrialization on air and water quality (Taagepera 1981). Opposition leader, Dainis ¯Ivans¯ would later recall that, “In Latvia everything began with the movement to save the environment,” claiming that was the moment, “the big awakening had started.” (Dreifelds 1996, p. 59). ¯Ivans’s¯ organization, the Latvian Popular Front ﬁrst emerged in this context of public protest and discontent. The PF quickly took on a coordinating role absorbing the many smaller organizations and in doing so became a general catch-all organization for all protests, whether historical, cultural, or environmental (Penikis 1996). In response, the USSR launched a rival organization, the Interfront. Interfront was to have branch organizations in all the Soviet Re- publics, and sought to persuade Soviet citizens that any form of nationalism or secessionism was reactionary. The main source of disagreement between the two “fronts” was over the power of the center to control the periphery: the PF demanded greater autonomy and recognition of Latvian sovereignty, while Interfront asserted the right of centralized communist control even as the USSR instituted moderate liberalization reforms. The debate between the two organizations increasingly took on an ethnic quality. Nationalists became a dominant force within the Popular Front which eventually became an overt pro-independence movement. Interfront, meanwhile increasingly dropped its “communist” identity for a “Russian” one. By the early 1990’s, Inter- front publications were articulating a “Russian idea”, rather than a Marxist one, and at one point even implored citizens to “Remember that you are Russian!” explicitly dropping the ideological platform for an ethnic one (Kolstø 1995, p.113-5). In this way, independence began to look like an ethnic issue, with popular perception putting ethnic Russians on the side of remaining in the Russian-dominated Soviet Union and ethnic Latvians on the side of an independent Latvian state. Nevertheless, the interpretation that these two blocs were perfectly ethnically divided is

164 inaccurate. It is true that nearly all ethnic Latvians supported independence from the Soviet Union, as advocated by the PF. It is also true that Interfront was supported almost exclusively by non-Latvians. However, it is not true that all non-Latvians supported Interfront, as many ethnic Russians supported the PF and independence. By 1991, an overwhelming 94% of ethnic Latvians supported independence from the Soviet Union, while 38% of non-ethnic Latvians did as well. Many ethnic Russians held high positions in the apparatus of the PF, most notably Anatolijs Gorbunovs4 who served as Chairman of the Latvian SSR’s Supreme Soviet, but nevertheless publicly broke with Moscow and lead a faction of high-ranking communist officials to the Popular Front. Gorbunovs’s role coordinating a link between the Popular Front and the communist party would earn him the position of Latvia’s first post-independence speaker of Parliament (Purs 2012, p. 85).5 Many ethnic Russians found it easy to support the PF because it was extremely ideologically diverse. While there were those who advocated for total independence from the Soviet Union and tribunals which would hold Soviet officials legally accountable for an illegal occupation of the country, there were also moderate wings who wanted little more than increased autonomy within a reorganized, decentralized Soviet system. Still others were motivated only by increased control over education and history, and accountability for environmental degradation. Questions of citizenship and the ethno-national nature of the Latvian state were generally tabled by the PF, which in the late 1980’s and early 1990’s promoted an extremely inclusive line and welcomed members of all ethnic groups and ideological affiliations (Bennich-Bjorkman 2017). The PF would be the most important organization supporting the Latvian independence referendum in 1991. Following the success of the referendum, its members would also guide the transition to an independent Latvian state. The most consequential decision made by PF leaders was not to create a new Latvian state, but rather to “restore” the constitution of 1922. Unlike many Soviet republics, Latvia and the other Baltic states had experienced their own statehood in the interwar period. Give the PF’s reliance on commemoration of that statehood in

4Gorbunovs is from a family of Russian Old Believers, so while he is technically ethnically Russian, he is in a diﬀerent category of most ethnic Russians in Latvia. 5In the early 1990’s, Gorbunovs’s stance on ethnic Russian rights was also decidedly out of step with the preferences of many ethnic Russians. In 1992, Gorbunovs advocated that citizenship issues should be decided by a referendum, with voting restricted to only the descendants of pre-1940 citizens (Lieven 1994, p. 305).

165 mobilizing its supporters, that period had come to be idolized by large sections of the Latvian population. The decision was therefore made to adopt a legal interpretation that reflected the historical narrative celebrated by those demonstrations and commemorations. They argued that Latvia’s incorporation into the Soviet Union had been illegal under international law, and that the only legal remedy to this essentially criminal act was to restore the Latvian state as it existed in 1940. This decision reestablished the electoral institutions which would facilitate an extremely volatile and unstable party system (Davies & Ozolins 1994). This conscious nostalgia for the interwar republic was politically popular with most of the former PF. As the interwar constitution required the newly formed parliament to appoint the state President, the first post-Soviet government chose Guntis Ulmanis, the grandson of the brother of Karlis¯ Ulmanis, the president of Latvia at the time of the Soviet annexation.6 The Latvian electoral system of 1922 also contained a unique form of open-list voting. Latvian voters must first cast their vote for a party list, but within that list they are also free to give each individual candidate on the ballot either a plus or a minus. Pluses move candidates up the list from their original position; minuses move them down. As a result, Latvian party leaders have less control over their candidate lists than leaders in other countries. An extremely unpopular party leader may find herself low enough on the list after an election that she is not able to enter parliament if she earns enough minuses, just as a candidate who was not popular with the leadership of a party may start in a low position on the party list but rise up through massive support among voters. While the system does empower individual candidates and encourage them to invest in their personal popularity, it nevertheless requires party membership. A candidate cannot run as an independent, so she must join a party in order to have a hope of ever being seated in parliament. Post-Soviet reformers did make one change to the interwar electoral law: implementing a 5% electoral threshold for all elections after 1995 (Smith-Sivertsen 2004). Cognizant of the chaos caused by the enormous number of parties in the interwar period, the threshold was intended to reduce the influence of fringe parties and result in more stable governments with fewer coalition partners.7 The threshold also

6Guntis Ulmanis had spent the decades before independence as a member of the communist party, and had been living under his step-father’s surname at the time. He changed his name back to “Ulmanis” after independence. 7Anecdotally, some candidates and MPs in Latvia complained in interviews about what the Latvian electoral system does to intra-party policy making. Many politicians decide to run for oﬃce on the basis of their own personal

166 encouraged candidates to come together to form political parties larger than the comparatively small factions that had existed in the independence movements. Small factions united by personal ties and single-issue positions often struggled to locate other would-be candidates whose interests and ideologies similar enough to put together a platform, and then form a pre-electoral coalition to pool support (Lieven 1994, Golder 2006). These historical and institutional factors are helpful in understanding why the Latvian political party system shows extreme volatility, not just in the immediate post-Soviet period, but even decades letter. The communist party faction represented by Interfront was rendered almost entirely irrelevant by early losses in founding elections. As a result, there is no single strong Soviet successor party in Latvia, with former communist leaders breaking apart to form several smaller parties and factions. As the Popular Front and its leadership in the late 1980’s had been most concerned with creating a large, “big tent” anti-Soviet political factions, they tended to ignore many issues that would eventually come to the fore. By focusing on issues of environmental protection and historical grievance the Popular Front effectively tabled issues of citizenship, political reform, and economic transition—the main issues that would eventually dominate electoral contests. The Popular Front was effectively a large collection of small, atomized factions, coordinated around a broadly anti-Soviet agenda. It was too internally diverse to survive as a political party, but the smaller factions were too small to be viable political parties on their own. As a result, elections are usually dominated by institutionalized coalitions which are rarely permanent. New parties, often composed of reconfigurations of existing elites, have emerged at every single electoral contest, and many of them win large portions of votes. Parties also sometimes form purely on the basis of personal connections between candidates. When this happens, those parties often do not outline comprehensive platforms or ideological positions, support, or popularity within a specific subset of the population. Often this popularity is large enough for them to get a large number of “plus” votes on the open ballot system, but not large enough to clear the 4% threshold for entry in Parliament. Those politicians therefore have strong incentives to join existing parties, and then encourage their voters not only to give them a plus on the open-list ballot, but to give minuses to every other candidate within the party, knowing that moving everyone else down the party list may be the only way to ensure their preferred candidate actually wins a seat. One established candidate from a mainstream party told me that they had started to feel that most of their campaigning energy was spent trying to fend off the negative consequences of minuses from new candidates within their own party, not convincing voters to support the party itself. They said that while they didn’t like the idea of telling supporters to cross out other members of the party, they started to feel the need to do so to preserve their own seat. While anecdotal, this link between electoral institutions and intra-party animosity may account for the extremely high rate of new party formation and party division in Latvia, even compared to other post-communist states (Kreuzer & Pettai 2003).

167 and instead campaign on the basis of the few interests or positions shared between the candidates. The result has been extremely high volatility with virtually no strong individual-level party identity to speak of. Nevertheless, a scholarly consensus has nevertheless emerged that two cleavages have dominated Latvian politics in the post-Soviet period: the economic divide between socialists and liberals, and the ethnic divide between Latvians and Russian-speaking Slavs (Pabriks & Stokenberga 2006, Zake 2002, Smith-Sivertsen 2004, Saarts 2011). The decision to restore the interwar constitution created an enormous divide between ethnic Latvians and ethnic Russians which persists as one of the most important political cleavages in the country almost three decades later. The 1922 constitution had been extremely liberal towards the rights of minorities in the country, and while the authoritarian Ulmanis regime had ultimately limited the linguistic and economic rights of non-Latvians, the constitution granted ethnic minorities the same civil liberties as everyone else. But the legal interpretation espoused by the transitional authority was that Russians who had migrated to Latvia were the remnants of an illegal occupation force. Russian Latvians were not citizens, but invaders. Russian-speaking Latvians were therefore not automatically citizens of the new, restored Latvian Republic, but instead foreigners whose rights to naturalize was a political matter to be decided by the new leadership (Brubaker 1992). Whereas Russian speakers had previously been welcomed as an important part of the Popular Front’s mobilization eﬀorts, some hardline politicians were now vocally opposing the Russian-speaking Latvians, as one prominent leader would say, “You are living illegally on the Latvian territory...you will be forced to leave sooner or later (Lieven 1994, p. 307).” This hardline nationalist faction would ultimately guide Latvian citizenship policy. The oﬃcial position enshrined in the law was that communist rule in Latvia had been an illegal occupation of the country by the Soviet Union, and the only legal constitution that could govern the transition process was that which was in place in 1940. At the same time, by this logic, the only legal citizens of Latvia were those who had been citizens in 1940, or would have inherited Latvian citizenship from their parents since then. Of course, this excluded the large numbers of residents of Latvia who had emigrated to the Republic from elsewhere in the Soviet Union and their descendants. Most ethnic Russians, even those who were born in Latvia and had never

168 known any other home, suddenly found themselves denied citizenship and excluded from political participation. For some hardline nationalists, the policy also resolved a difficult dilemma of the ethnic composition of the Latvian state. Of all the three Baltic post-Soviet states, Latvia was where the titular majority was the most vulnerable to becoming the minority. In 1994, the year the law on citizenship was passed, ethnic Latvians comprised only 54.2% of all residents (Chinn & Truex 1996). Nationalist leaders predicted that if all residents of the country were granted citizenship in a democratic state, the non-Latvian population would be politically strong enough to undermine the Latvian nation-state. But this somewhat draconian nationality policy would ultimately create two different fronts of conflict for the new Latvian state: one between the Lat- vian nationalists and the disenfranchised ethnic minority population, and another between the Latvian state and the western Europe, which saw the citizenship policy as a violation of both fundamental democratic values, and existing human rights laws as established by the United Nations and the European Union (Chinn & Truex 1996).8 The citizenship laws put Latvian nationalists in a difficult position. On the one hand, they needed to ensure the continued ethnic Latvian majority if the state was to remain ethnically and culturally Latvian in nature. They also needed strong ties to the west through NATO and the European Union in order to protect their sovereignty from what they saw as their main geopolitical threat: Russia. But policies guaranteeing ethnic Latvian dominance put the country at odds with European standards on human rights and democratization. The Copenhagen Crite- ria, which outlines the standards applicant countries must meet before joining the European Union, specifically describes the protection of minorities and the extension of full civil liberties to minority populations as a requisite component of democracy (Hillion 2004). Ironically Lat- vian politicians often found European leaders sympathetic with the plight of ethnic Russians. European officials pressured Latvia throughout the 1990’s to increase the rate of naturalization and refused to accommodate initial nationalist plans for a program encouraging voluntary “repatriation” of ethnic Russians to the Russian Federation. Under pressure from the European Commission and the OSCE, Latvia abolished quotas for naturalization, reduced costs and requirements to naturalization, and established exemptions to certain requirements on vulnerable

8For a more comprehensive discussion of Latvian citizenship policy, see Commercio (2008), and Agarin (2010).

169 populations deemed disproportionately onerous (Morris 2003). Latvia’s minority citizenship issues were often handed through the OSCE. European leaders in the early 1990’s were more fundamentally concerned with security and stability than encouraging full democratic citizenship and human rights, and often treated de-militarization as this more pressing issue in the immediate post-Soviet period (Vrbek 2015). While most Soviet troops had left Latvia by 1994, some units of the Russian armed forces remained scattered throughout Latvia until 1999. Russia had sought to militarize the issue of the Russian minority in Latvia by making minority rights a condition of troop withdrawal, and the OSCE made delinking the two a very high priority in order to avoid potential escalation and violence (Muiznieks 2011). The OSCE therefore put pressure on Latvia to resolve the citizenship issue as a way of neutralizing potential security threats to Latvia from Russian claims to protect ethnic Russians in the Baltic. This pressure from European institutions is correlated with an increase in applications. As seen in Figure 6.1, naturalizations increased markedly in the late 1990s and early 2000’s, when Latvia implemented policy reforms in support of its application to join the European Union was at its highest. Naturalizations surged again following Latvia’s accession in 2004. Applications have tapered oﬀ in recent years, largely indicative of the fact that those remaining non-citizens in Latvia are not applying for citizenship either because they are unable, or unwilling (Kuczyńska-Zonik 2017). Thus while the bulk of Latvia’s non-citizens have gained citizenship over the past two decades, citizenship policy in Latvia has been vitally important in shaping the relationship between the two ethnic groups. Even though many non-citizens were born in Latvia and have known no other home, having a parent or grandparent who had immigrated to Latvia as a result of Soviet policies relegated many ethnic Russians to second-class status. Many ethnic Russians found themselves deprived of voting rights and other civil liberties as a result of historical events which happened before they were born, and factors over which they had no control.

170 Naturalizations in Latvia 20000

15000

10000

5000

1000 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019 Year

Figure 6.1: Naturalizations in Latvia: 1995-2018

6.3 The Political Use of Ethnicity in Latvian Elections

In this section, I explore how Latvian political actors use this history and the diﬀerences between ethnic communities to get elected. Latvian politicians have a unique, state-sponsored resource that their Bosnian counterparts lack. In Latvia, politicians hoping to get on the ballot must join a party list, and that party must submit a political party platform. These pre-election programs are not only part of the public record, but are distributed to media to be published in newspapers and broadcast on TV and radio, and also provided to voters on election day when they receive their ballots (Ikstens 2017). As a result, these programs are not just a bureaucratic hoop that parties and candidates must jump through, but are eﬀectively state-sponsored free advertising. They are the only legal way that a party can communicate to voters when they are actually in the voting booth, and parties have strong incentives to portray themselves in the best possible light. These programs are limited by law to 4,000 characters, meaning that Latvian political programs are extremely short by international standards.9 This was by design: shorter party pro-

9For comparative purposes, the German Christian Democratic Union’s 2017 party program was over 150,000

171 grams are more accessible to everyday citizens. Four thousand characters can easily be printed on a pamphlet, and read relatively quickly by a typical voter, and Latvian citizens are far more likely to have seen and considered official party programs than their counterparts in other democratic countries. The character limit also ensures comparability across parties for the purposes of comparative research, as it forces all parties, regardless of electoral viability or professional- ization, to identify only their most important campaign themes and present them to voters. In this section, I discuss party naming conventions and the rhetoric parties use regarding ethnic identity, drawing on the party platforms that parties have submitted when registering to contest elections. I find that most often ethnic affiliation is conveyed through reference to language, culture, and the differences in historical legacies which represent the most salient differences between the two groups. Latvian ethnic affiliation is communicated through an emphasis on a history of colonial occupation, mostly during the Soviet period, but references are also made to the pre-modern domination of Latvia by Germans and Russians. Politicians campaigning on ethnic Latvian platforms often speak of the vulnerability of the ethnic Latvian community and of the risk of group extinction as a reason to elect ethnic representatives. Ethnic Russian parties also campaign on the basis of representing a vulnerable minority. They often propose a counter argument to the historically-justified claims of ethnic Latvian parties, arguing that the achievements of the Soviet Union and the Russian-speaking population it brought to the country are worthy of celebration on their own rights. A third, but much smaller group explicitly campaigns on the grounds of civic nationalism. These parties argue that the Latvian state should champion no single ethnic group, and should instead advocate for cultural pluralism, multilingualism, and policies which incorporate all members of society, regardless of ethnic background, as full democratic citizens. A fourth group of parties completely ignores the ethnic divide of the country to instead campaign entirely on ideological or policy positions. characters. The American Republican Party’s platform for the 2016 election was over 240,000 characters. The anti- establishment Five Star Movement in Italy had even more than that in its 2018 platform with well over 350,000 characters.

172 6.3.1 Party Names

As in Bosnia and other countries, party names are important signals of ethnic aﬃliation. In Latvia, several parties include the words “Latvian” or “Russian” in the name of the party. Par- ties also use names which employ references to Latvian history, appropriating the names of independence movements or even pre-Soviet political parties. Other parties use naming conventions to distance themselves from ethnic labels and nationalist positions by adopting names which emphasize civic nationalism, internationalism, and universal human rights. Given the relative importance of historical commemoration during the independence movement in Latvia, it is perhaps unsurprising that many parties adopt names that have historical signiﬁcance. References to pro-independence and anti-Soviet names appear frequently. For the 1995 elections, politicians who had been part of the Popular Front attempted to rebrand as a political party, also called the “Latvian Popular Front.” The name was clearly meant to highlight the pro-independence contributions its members had made in the late 1990s, and the party adopted a more hardline ethno-nationalist stance. Voters clearly found the appeal unconvinc- ing, and the party failed to clear the electoral threshold. Many of its candidates rebranded, and competed as a Christian Democrat party in the following election. The same could be said of Latvian Way, a party which also contested elections in 1995 on an ethno-nationalist platform.

The Latvian name for the party—“Latvijas Cel,š”— uses the same word for “way” as the “Baltic

Chain” (“Baltijas cel,š”) discussed above. The event is considered an important milestone in the history of the collapse of the Soviet Union. Latvian Way was far more successful electorally than the Latvian Popular Front, participating in governing coalitions following the 1993, 1995, and 1998 elections. A third example of this phenomenon would be the party “Helsinki-86,” which formed not in 1986, but in 1998, and sought to build on the legacy of the eponymous organization (which was formed in 1986) and was one of the earliest instances of anti-communist mobilization in Latvia. Helsinki-86 also took one of the strongest pro-ethnic Latvian positions of all the parties, arguing not only for linguistic and cultural rights for ethnic Latvians, but also for the establishment of a cabinet-level “Ministry of Repatriation,” which would coordinate the “return” of ethnic Russians back to the Russian Federation. All three of these parties are named after independence-era movements which were not

173 ostensibly ethnic at the time. The original Latvian Popular Front had large numbers of ethnic Russians, and as discussed above purposefully reached out to ethnic minorities and Russian-speaking Latvians to create the broadest coalition possible in support of their pro- independence movement. Likewise, the Baltic Chain protest was explicitly designed as a demonstration of unity, comprising not only ethnic Latvians, but also Lithuanians, Estonians, and pro-independence Baltic Russians (Eglitis & Ardava 2012). The leadership of the original Helsinki-86 organization was comprised entirely of ethnic Latvians, but was nevertheless mobilized around a treaty which advocated basic human and civil rights. It is a great irony that a party named after a treaty which demands “...the effective exercise of civil, political, economic, social, cultural and other rights and freedoms all of which derive from the inherent dignity of the human person...” and “...conformity with the purposes and principles of the Charter of the United Nations and with the Universal Declaration of Human Rights” (Helsinki Declaration 1975, Ar- ticle VII), would advocate the forced expulsion of those born in Latvia—denial of the freedom of movement and residence established under Article 13 of the Universal Declaration of Hu- man Rights. Instead, these parties claimed to be continuing a new form of the independence movement, even after Latvia became a sovereign state. No longer focused on the struggle for independence, these parties were now fighting for the sovereignty of the Latvian people, and the dominance of the Latvian ethnic community within an ethnic state. For these parties, the legacy of the independence movement is not necessarily one of human rights or anti-communism, but it was a struggle for the rights of ethnic Latvians against ethnic others. Appropriating the names of the movements signals to voters that these parties “continuing” that fight, even if that particular struggle was but a small component of the original organization’s agenda. Other parties show their connection to the ethnic Latvian communities by adopting names that show a pre-Soviet historical identity. The clearest example of this is the Latvian Farmers’ Union, whose name suggests not only an agrarian link and a connection to the rural population outside the major cities, but also a link to the inter-war period when farmers were an important force. The eponymous Latvian Farmers’ Union, formed in 1917 but suppressed during the Soviet period, was the second-largest party in Latvia in the election of 1931, and the most important conservative force in Latvian politics at the time. By adopting the name of a party

174 that hadn’t existed for more than half a century, party leaders conveyed to voters that they were in effect reviving an older, established political party in the same way that independent Latvian statehood was also being resurrected after decades of illegal Soviet occupation. In its 1995 party program, the party explicitly claimed the heritage of that party, noting, “The Latvian Farmers’ Union, historically established and grown up along with the Latvian nation, is the largest party in Latvia. It is the party of Karlis¯ Ulmanis,10 Zigfr¯ıds Meierovics,11 Janis¯ Čakste,12 and Guntis Ulmanis.”13 In this way, the choice of name seems coordinated with an attempt to convince voters that the party is a continuation of the party most responsible for independent Latvian nation-state. The party, like many in Latvia, has evolved throughout the post-Soviet period through a series of mergers, but in its current incarnation as part of the Union of Greens and Farmers remains an important party in Latvia, having been included in governing coalitions and frequently holding the presidency. Another important way parties use their names to show an ethnic affiliation is to exploit a quirk in the Latvian language that differentiates between the Latvian state, and the Latvian language or ethnic identity. Many parties identify themselves using “Latvijas,” the genitive sin- gular noun form of the word “Latvija,” meaning either the territory or the state of Latvia. Other parties identify themselves as “latviešu,” the genitive plural form of “latvietis,” which refers to an ethnic Latvian or Lettonian person. The difference is difficult to render in English, but very clear in Latvian. A “Latvijas” organization is simply an organization that exists within the country of Latvia. A “latviešu” organization is one that is filled with ethnic Latvians, or reflective of ethnic Latvian culture, or speaks the Latvian language.14 Political parties use this difference in their names to convey their ethnic affiliation. This is perhaps most clear in the case of “LATVIEŠU LATVIJA!” an all-capitals party name which could be (awkwardly) translated as “An Ethnically Latvian Latvia!” or, “Latvian-speaking Latvia!”, but also seen in “Latviešu

10First Prime Minister of Latvia, and leader of the interwar authoritarian state. 11Interwar Latvian Foreign Minister, celebrated for negotiating recognition of Latvian statehood and support against Bolshevism from western powers. 12First President of Latvia. 13Grand-nephew of Karlis¯ Ulmanis, appointed as the first post-Soviet president of Latvia. 14The difference also exists in Russian: note the difference between “русский” and “российский”, and “ла- тышский” and “латвийский”. Perhaps the best way to understand the difference is to look at the Latvian Russian Union. Composed almost entirely of ethnic Russians, the party calls itself “Latvijas krievu savien¯ıba”—literally “Latvia’s Russian People’s Union.” A “Latviešu krievu savien¯ıba” is almost a logical impossibility in this context, since such people would simultaneously be ethnically Russian and ethnically Latvian.

175 partija,” which would normally be translated as “Latvian Party,” but actually means “Party of Ethnic Latvians” or “Party of Latvian-Speakers.” Perhaps the most perplexing outcome of this linguistic quirk for English speakers is the election of 1995, which saw both a “Latviešu Zem- nieku Savien¯ıba” and a “Latvijas Zemnieku Savien¯ıba” competing against each other. Both can be translated as “Latvian Farmers’ Union”, but the former is an ethnic designation (i.e., the farmers are ethnic Latvian) while the latter is a territorial designation (i.e., the farmers live in Latvia). For those parties which explicitly campaign as champions of ethnic Russian causes, the naming is often extremely straightforward: they identify themselves as “Russian” in the name. There can be little doubt as to which ethnic group the “Russian Party” claims to represent, likewise with the “Latvian Russian Union,” or the “Russian Citizens of Latvia Party.” Perhaps the most subtle reference to Russian identity comes from the Dzimtene Patriotic Political Association. “Dz- imtene” can be translated as “homeland.” Latvian nationalist monuments, including the famous Freedom Monument in central Riga often declare their loyalty to the homeland using the word “tevzemei.”¯ “Tevzemei”¯ comes from the word “tevu”¯ meaning “father,” and “zeme,” meaning land. It is a literal, word-for-word translation from the German “vaterland.” “Dzimtene,” on the other hand, is modiﬁed version of “dzimt,” to be born. It is a literal translation of the Russian “родина”— “motherland” or “land of birth.” Given Dzimtene’s expressed position that, “There is one dzimtene, one country, where Latvians, Latgalians, Russians, Ukrainians, Poles, Ger- mans, Byelorussians, Jews, Roma, etc. live together. We need to ﬁnd the strength to be united,” this alternative form of “homeland” presents an alternative to the German-derived word which they saw as appropriated by ethnic Latvian parties. The party formed to contest the 2006 election, and was comprised primarily of ethnic Russian candidates who left other parties as they drifted more towards ethnic Latvian positions. It failed to clear the threshold for representation that year, and did not compete again. Many ethnic Russian parties adopt names which stress the value of human rights. The party “For Human Rights in a United Latvia,” which has contested in several electoral cycles, formed from the merger of several separate Russian-dominated parties. In 2002 the party campaigned on a broad leftist platform of anti-privatization and increased pensions and wages for state work-

176 ers, as well as directly attacking nationalist rivals. The party also made its ethnic affiliation clear, arguing that Russian should be given official language status, and that naturalization for Rus- sian non-citizens should be made easier. Over time, the party’s position on Russian language rights became more extreme, and by 2011, the party was campaigning for higher education opportunities in more international languages like Russian and English (i.e., not Latvian), and for opening the border with Russian and Belarus in order to normalize trade and increase ease of personal travel. They claimed that officials should “speak the language of the taxpayer’—a rebuke to Latvia’s monolingual policy of Latvian as the only official state language. The party was highly successful in mobilizing Russophone support, earning second place with a quarter of the seats in parliament in 2002, but eventually declined as factions within the party split to compete independently. The other most common identifier in a party’s name is the word “Harmony.”15 As the largest party in Latvia today, Harmony Center has built its support by putting together a coalition of Russians and Russian speakers with ethnic Latvian leftists. The term “Harmony” first appeared on the Latvian ballot as the “Peoples’ Harmony Party,” when it campaigned on a platform of Russian language rights, extension of citizenship to Russian-speaking non-citizens, and a repatriation law that would encourage ethnic Russians who had emigrated from Latvia to Rus- sia to return to their homes. Since then, various forms of “Harmony” have appeared as the party reformed, rebranded, and evolved. It went from the “Peoples’ Harmony Party” in 1998 to “Harmony Center” in 2006, then the “Harmony Social Democratic Party” in 2014. The term “Harmony” is meant to convey compromise and the peaceful coexistence of multiple groups in Latvia. The party has varied in the degree to which it directly appeals to Russian ethnic interests, but the party has always been supported predominantly by Russian speakers, and has always been opposed to an ethno-national state dominated by ethnic Latvians. In short, both Latvian and Russian groups use their names to convey to voters the ethnic affiliation of their party. Ethnic designations of “Latvian” and “Russian” appear directly in the

15 “Saskan, a” in Latvian, and “согласие” in Russian. Today’s Harmony party officially translates the word into English as “Concord,” and every Harmony supporter and elected official I spoke with during fieldwork was personally unhappy that the party has come to be known as “Harmony” abroad, noting that the words “harmonija” and “гармония’’ also exist in Latvian and Russian, and are not the official name of the party. In my own interviews,

even when speaking English or Russian, most people referred to it as “Saskan, a.”

177 names of several political parties, helping voters immediately see the ethnic aﬃliation of the party when making voting decisions. Ethnic Latvian parties also show the tendency to rely on historical markers to convey their connection to ethnic Latvian nationalism, and the independence of the Latvian state from foreign interference. While some of the historical signiﬁers the Latvian parties adopt could be plausibly construed to include non-ethnic Latvians, ethnic Rus- sian parties do not show the same willingness to adopt historical allusions in their party names. Instead, Russian parties tend to rely on labels that stress universal human rights or multi-ethnic coexistence.

6.3.2 Campaign rhetoric

Of course, voters do not rely solely on party names when making voting decisions, and the positions that a party takes as well are important. In Latvia, parties convey their ethnic affiliation through a variety of policy positions and statements of their own identity. Latvian political parties often employ historical allusion to convey an ethnic identity to their voters. These parties differentiate between “us” and “them” by emphasizing historical grievances against former non- Latvian occupiers, and argue that political representation for coethnics may help redress group grievances. They also often stress the historical differences between Latvians and non-Latvians even if many of those distinctions are less relevant now than they once were. These parties praise rural and agrarian ways of life in a reference to the fact that ethnic Latvians were often relegated to agricultural production while both German and Russian colonizers were more likely to settle in cities. Latvian parties also tend to stress the vulnerability of the Latvian nation given its small size and geopolitical position between great powers. They also make reference to the citizenship laws in Latvia which almost exclusively disenfranchised Russian-speaking residents, and promise linking benefits to citizenship as a means of diverting benefits to coethnic supporters. Perhaps the most straightforward way that parties identify their affiliation is to openly declare that they will enact parties that benefit coethnics. The “Latvian Latvia” party mentioned above promised to represent the ethnic interests of Latvians, proclaiming that “the national economic, capital, budgetary, taxation, and monetary policy all must serve the interests of Latvian

178 people. The Latvians will have jobs and bread!” Likewise, the National Strength Society declared in 2006 that, “the economic power of Latvia must belong to Latvian people. Products made in Latvia should be given priority in the Latvian market.” This economic nationalism which promises that coethnics will benefit from state policy is a clear indication that the party sees itself as championing ethnic interests. In 2002, the New Era Party proclaimed that, “Latvia is the only country in the world with the responsibility for preserving the Latvian people and the Latvian language.” While New Era would ultimately win the largest share of seats in that year’s Parliamentary election, the sentiment perfectly captures the feeling of many Latvian ethnic parties. For these parties, history has shown that the Latvian people are vulnerable. They are a small group, split between the two much larger political and cultural forces of the Germans and the Slavs. The Latvian people need the protection of a state of their own, and that state must take special care and grant special privileges to ethnic Latvian identity, if for no other reason than that no other country is likely to do so. The National Democratic Party similarly has argued that “God has given Latvia to the Latvians. Latvia is a nation-state, which must ensure the survival and prosperity of the Lat- vian people. Latvians have the right to laws and citizenship policies which protect the national character of the state.” For many parties, the hallmark of the Latvian identity is the Latvian language. Not only is it grammatically distinct and mutually unintelligible from other European languages (with the exception of Lithuanian), it has also long been subjugated and relegated to second-class status to German, Russian, or—in the post-cold war era—English. It is hard to overstate the importance of language to ethnic Latvian parties. Neither the Teutonic knights who first Christianized Latvia nor the communist bureaucrats who administered the Latvian SSR made serious efforts to learn Latvian in large numbers, and as such the Latvian language is perhaps the most reliable marker of Latvian ethnic identity. For many Latvians it is also symbolically important as a sign of sovereignty, respect, and self-determination on the global stage. Many parties therefore make the Latvian language central in their outreach efforts to voters. While large numbers of parties explicitly state that Latvian is the only language that will be spoken in Latvia in somewhat abstract terms, there is some variation in the intensity of the sentiment. Several parties

179 simply stated that Latvian is the language of Latvia, and should have official status as the language of state. Helsinki-86 stated that its main policy priority was the “defense of Latvian Law, Language, and Culture,” suggesting not only that those things were important, but also that they might somehow be related. In this interpretation the survival of the ethnic Latvian people and the Latvian language is contingent on an ethnically Latvian political authority capable of protecting them. The Popular Front in the 1995 elections held that the vitality of the Latvian language should be a policy priority for the state, arguing that, “its most important manifestation is national culture and a fully-vibrant Latvian language.” Zatlers Reform Party, a party formed in 2011 by a former President, echoed this sentiment by saying, “let’s foster and promote Latvian culture and the Latvian language as the most important reason for the existence of the Republic of Latvia.” Others go further arguing that state policies should also encourage the exclusive use of Latvian in the media or in private. The Samnieks Democratic Party16 stated that “Latvian will be the only language of state, family, and society.” This position was likely not well-received in the very large number of families which speak Russian or other languages, but was popular enough for the party to come in first in the 1995 election. Other parties offer more than relatively vague statements of valuing the language, but concrete policies, meant to defend the Latvian language and ensure its dominance. The most common of these is in the field of education. Many parties in the post-Soviet period have proposed that minority-language education should not be allowed, requiring that students not only study Latvian, but study all subjects in Latvian. Many parties have proposed that education in the broadest sense should be only in Latvian, while others are more specific. The above-mentioned National Democratic party argued that all state-provided free education should be only in Lat- vian, and that university education which is not free should also be subsidized by the state for the most talented Latvian-speaking students. All for Latvia! proposed that any educational institution funded by the state should be allowed to speak only in Latvian and admit only Latvian- speaking students, and that the only foreign languages taught in schools should be those of

16“Samnieks” is hard to translate into English. It traditionally meant “owner” or “landlord,” and was the term historically used to describe wealthy landowning aristocrats. But it also means “boss,” “manager,” or “supervisor.” It can also mean “host,” both in the sense of someone who invites and entertains guests, and the biological sense of an organism oﬀ of which a parasite feeds. In this political context, it is likely meant to convey a sense of being “savas zemes samnieks,” or “lord of your own land.”

180 the European Union—a standard which would notably exclude Russian and Ukrainian. Father- land and Freedom, a far-right nationalist party, has explicitly argued that students in minority- dominated areas should be required to learn only in Latvian. Other common policy proposals concern media. The National Alliance promised their support for policies which would require the use of Latvian as the national language in all public media, as well as in business and private enterprise. Even Unity—a party frequently portrayed as a centrist party that does not campaign on ethnic interests—promised their support in 2010 for “cultural space, quality public media, and adequately supported culture in the Latvian language,” as well as, “Latvian language protection, a language development program, and civic education aimed at consolidating society.” Fatherland and Freedom expanded the domain of language laws to include the administration and civil service, proposing that any state employee be required to submit to regular language proficiency tests as a way of ensuring linguistic purity among civil servants. This policy may have been prompted by the fear that ethnic Russians were becoming proficient in Latvian just long enough to pass the language tests required of the naturalization process, but still speaking Russian at home, thus contributing to the “corruption” of the Latvian language. In some of these proposals there is ambiguity as whether this is necessarily ethnically exclusive. After all, it is entirely possible to be a civic nationalist, but still believe that it is advantageous for all members of society regardless of ethnicity to be able to communicate with each other. But many parties make it clear that their intent is to limit opportunities for Russian speakers. The National Alliance promised to, “recruit workers in the service sector who have no Russian language skills, as a way of preventing discrimination against Latvian youth in the labor market.” All for Latvia! also reminded voters in 2011 of its previous achievements in proposing laws that prevent employers from listing Russian language skills as a job requirement in order to prevent Russians from having advantages over Latvian job-seekers. It seems unlikely that preferring workers who speak fewer languages over those who speak more is in anyway an attempt to increase economic efficiency or competitiveness, and can only be a policy of favoritism for ethnic Latvians. Several parties identify their ethnic nature by specifically mentioning a threat to the Latvian

181 nation, and promising to respond to it. One of the most prominent examples of this is demographic crisis. Given the high rates of immigration from Russia and the rest of the Soviet Union, Latvia became the former Soviet republic where Russians and titular nationals were closest to parity, making the possibility of Russians becoming an ethnic majority very real in Latvia. Many parties explicitly address this demographic crisis, such as the Democratic Party, which in 1998 claimed that, “The Latvian people are in the most critical demographic situation of their survival, and are losing many of their reasons for living,” and argued that the state must take up the role of the “patriotic guardian” and “ensure the material equality of the nation through an assessment of its demographic future.” The Latvian Social Democratic Union has promised to “stop the extinction of our people,” and the Our Country party went so far in 2002 as to claim, “Latvia is the home of the Latvians, given to them by God, and it is now impossible to stop their extinction without extraordinary means.” Some parties explicitly argue that actions should be taken that would fix the demographic proportions of the population and preserve the ethnic Lat- vian majority. The Fatherland Union in 2006 argued that, “Economic emigration must be halted, as it can create an influx of migrant workers to Latvia,” which they felt would jeopardize the fragile majority of ethnic Latvians. All for Latvia in 2006 argued that ethnic Latvians had previously been the victims of genocide, and as such special measures should be taken to increase their numbers.17 In the early post-Soviet period Fatherland and Freedom proposed the creation of a state fund which would financially reward ethnic Latvian families who had extremely large families. Several parties also advance proposals that would limit the rights of non-citizens, as a way of signaling their intentions to provide benefits to ethnic Latvians at the expense of ethnic outsiders who had already been deprived of citizenship. Helsinki-86 proposed that only citizens should be allowed to serve in the Interior Ministry, effectively depriving non-citizens of lucrative employ-

17What, exactly, those measures were remains unclear. However, on a trip to Riga in 2015, an oﬃcial from a right-wing ethnic Latvian party informed me that they personally thought outlawing abortion would be an important policy goal for Latvian nationalists in the future. This particular oﬃcial saw very little that was persuasive in the right-to-life arguments common in the abortion debate in the United States, but instead saw abortion rights entirely as a demographic issue. In this person’s mind, Latvian women were likely to terminate their pregnancy for the same economic reasons that ethnic Latvians would emigrate to other countries in the European Union: unemployment, low wages, and inadequate social services. Both emigration and abortion were causes of demographic decline of ethnic Latvians which could lead to a Russian majority in the country, and as such they felt it was the role of the state to seek to stop both of them.

182 ment opportunities, but also making it clear that ethnic Latvians would not likely to be subject to the authority of ethnic outsiders in the security and police services. The party also proposed legislating that only citizens could receive pensions or unemployment benefits, and be permitted to own land within Latvia. Other parties also proposed limits on non-citizen property ownership, including A Latvian Latvia, The Latvian Party, The Latvian Recovery Party, Our Land, the National Democratic Party, and the Anti-Communist Alliance. Proposals to limit property and asset ownership were especially popular in the mid 1990’s, when Latvia debated how to privatize many of the state-owned enterprises and assets in the transition from state-controlled communism to free market liberalism. Limiting ownership to citizens was an effective way to ensure that many non-ethnic Latvians would be ineligible to gain in the distribution of state assets, and thus excluded from the most lucrative opportunities of the transition to capitalism. Many of these parties argued that this denial of property rights was justified in light of their strong positions on the continuity with the pre-communist Latvian constitution. This narrative holds that in 1922, ethnic Latvians were rightfully given their own state, which was wrong- fully and illegally stolen from them by Soviet violence. Ethnic outsiders are therefore mostly migrants whose presence is only a consequence of that illegal occupation, and depriving those non-Latvians of their rights is morally and legally justified. In keeping with this trend in a slightly more subtle way, several parties specifically emphasize the rights of “citizens,” not “residents.” The Latvian Independence Party in 1995 stressed that it would defend the political rights of citizens under international law. This should not be interpreted as a defense of the rule of law, but instead a way of highlighting that the party draws an important distinction between the rights of citizens and non-citizens. Likewise, Fatherland and Freedom once announced their intention to provide quality healthcare for all Latvian citizens. Playing on anti-Russian sentiment which saw the Soviets as criminals and colonizers who destroyed and exploited Latvia, the party proposed to fund these generous benefits through confiscated smuggled goods and fines levied on those who polluted the environment. Parties also signal to voters their ethnic Latvian affiliation through their position on the continuity with the state as it existed in 1940. The National Conservative Party, joining with the far-right nationalist LNNK mentioned in their program in 1995 that not only did they support the

183 1922 constitution, but that they believed it specifically established a state exclusively for ethnic Latvians. Many parties, including the rightist Fatherland Coalition, All for Latvia! and Helsinki- 86 also stressed that they supported the 1922 constitution, mentioning that under international law the pre-war document should be the basis of citizenship policy and political negotiations with Russia. Of course, such a document would serve to justify that citizenship laws which were deeply unpopular with the ethnic Russian and Russophone population in Latvia. Many parties endorsed the 1922 constitution by referring to the Soviet period as an illegal occupation while stressing the negative consequences of Soviet rule on the ethnic Latvians. The Fatherland and Freedom coalition in 2006 promised a repatriation bill that would “eliminate the consequences of occupation and colonization” in Latvia. Given that in the same document the party declared its main priority was protecting the interests of Latvians it is hard not to understand that those “consequences” are likely the hundreds of thousands of ethnic Russians living in Latvia. Other parties also use this promise to restore the constitution of 1922 and to “decolonize” Latvia to propose policies which clearly privileges the ethnic Latvian community at the expense of others. Helsinki-86 in 1998 promised to restore Latvian sovereignty partially through guaranteeing jobs to the victims of the communist regime (i.e., employment for ethnic Latvians but not Russians), just as the National Power Coalition said that there was no place in Latvia for “alien and hostile powers that offend the Latvian people’s honor.” This position was accompanied by a promise to undo the effects of Soviet rule partially by destroying Soviet monuments and memorials in a bid to restore the true cultural and historical heritage of the Latvian people. Until 2006, parties were also mentioning Abrene in their party platforms as a major policy point. In 1946, the eastern part of Abrene county was transferred from the Latvian SSR to the Russian SFSR. Following the dissolution of the Soviet Union, Abrene remained part of the new Russian Federation, but many Latvians argued that since the 1922 constitution should be the basis of bilateral relations with Russia, Abrene should legally be returned to Latvia to restore the Latvian territory as it had existed in 1940. In 2007 an agreement was signed between Latvia and Russia maintaining the status quo without reference to the pre-Soviet borders, but some parties have used the Abrene issue to show their willingness to challenge Russia in order to maintain their adherence to the 1922 constitution.

184 Many parties representing ethnic Latvians also emphasize their connections to rural communities, and their desire to help rural populations within Latvia. While this in itself is not inherently representative of any single ethnic group, many ethnic parties seem to associate “Lat- vianess” with a rural identity. When Latvia First and Latvian Way ran as a united party in 2006, their platform equated their rural identity with a traditional Latvian culture that was agrarian, Christian, and socially conservative. That year, the party stated that, “we support using EU funds for the public interest, especially agriculture, forestry and fisheries. We will promote the development of new industries in the countryside—especially biofuel production.” The party then went on to say that they will “invest in the Latvian language, and the values and traditions of Latvian national culture.” This pattern is repeated in several parties, such as the Zigerist Party, which promised to maintain and protect Latvia’s existing state forests, as well as turn Latvia into a “successful farming country” though rural fuel and loan subsidies in 1998. This was coupled with a position that “The National Policy in Latvia must provide the Latvian people with opportunities for survival and prosperity in their own homeland. We will prevent the rapid extinction of the nation. We will respect and defend the Latvian language and the history of our people. Let the youth regain faith in their country, in the worship of God.” In this way, the party appeals to an image of the Latvian ethnic group that is both rural and endangered, promising to develop national policies that will preserve the culture and demographic position of a specific people. This is far from a unique combination of policies, as other parties which have taken strong positions on Latvian ethnic, linguistic, and cultural dominance—such as Helsinki-86, the National Alliance, Fatherland and Freedom, and All for Latvia!—have also proposed rural development money, agricultural subsidies, and increased funding for rural schools. This is not to say that all parties that take positions on rural areas are ethnic Latvian. The Freedom from Fear, Hate, and Anger party went so far as to take an overt anti-urbanization platform in 2014, proposing that as many Latvians as possible should live in the countryside, and even produce their own food if feasible. In that same platform it proposed that “Differences between nationalities and languages are no obstacle to national unity. Latvia was occupied, but today it is not occupied by occupiers. Everyone born after August 21, 1991 is entitled to Latvian

185 citizenship.”18 Harmony has also taken up pro rural-positions in its platforms over the years, promising development plans for agriculture in peripheral regions outside the capital city. Yet overall, rural and traditional identities largely remain the purview of Latvian ethnic parties. Unsurprisingly, there are substantially fewer ethnic Russian parties than ethnic Latvian parties. Russians and Russian speakers are the minority in Latvia, meaning that, by definition, there are fewer ethnic Russian voters than ethnic Latvian voters. Compounding the problem are the citizenship laws which almost exclusively disenfranchised ethnic Russians. Nevertheless, political parties appealing directly to ethnic Russian voters have been present in Latvian politics ever since independence from the Soviet Union. The most straightforward way that parties identify their ethnic affiliation to voters is to explicitly state so. While there can be relatively little doubt about which group the Russian Party appeals to, the official party platform makes it very clear that, “[t]he Russian Party is the party of patriotic Russian people living in Latvia. The defining feature of our party is the protection of the interests of Russian/Slavic citizens and non-citizens in Latvia.... Our purpose is to obtain rights for the Russian minority.” Separately, the Latvian Russian Union has explicitly stated that its goals were to “defend the interests of Latvian Russians,” and sought “to ensure the representation of the Latvian Russian community in parliament.” While the Russian affiliation of For Human Rights in a United Latvia might not have been immediately obvious from its name, its program makes it abundantly clear which side it is on. The party states that it is a human rights organization, precisely because it is a part of the Russian community which is most in need of human rights protection: “ForHRUL, as the party of the Russian community in Latvia, is the guarantor of equality and social justice among nationalities.” In this way, some parties argue that they should be elected to positions of power precisely because of their ethnic identity. The Latvian Russian Citizen’s party argued that Latvian political institutions privileged ethnic Latvians and facilitated their election into parliament, and only voting for ethnic Russians would serve to offset that tendency. In 1995 they took the somewhat radical position that, “Latvian Russians owe nothing to no one, certainly not for 50 years of

18It should be noted that this rural cosmopolitanism is not popular with voters, and the party should be considered marginal at best in Latvian politics. The party competed in the 2011 and 2014 elections, but never received more than 2500 votes, well under the threshold for representation in parliament.

186 Soviet rule...” and that the party would “...nominate our own Russian candidates as a means to promote societal cohesion.” This perspective is still present in Latvian politics decades later; as recently as 2014, the Latvian Russian Union maintained that its main goal was to, “ensure representation of the Latvian Russian community in the parliament.” These parties explicitly argue that ethnic Russians have particular interests separate from those of Latvians, and that those interests can only be served by electing co-ethnics. This ethnic identification often remains abstract, an expression of value and principle rather than policy proposal. However, some parties explicitly identify policy areas where Russians are disproportionately affected, and as such ethnic Russian representation is necessary in order to enact policies that can benefit the group. The most common area where this happens is language policy. Just as ethnic Latvian parties frequently portray the Latvian language as endangered and vulnerable to extinction given the small size of the Latvian population, ethnic Russian parties portray Russian as vulnerable to oppression and subjugation due to its minority status. Russian ethnic parties frequently argue that Russian should be given official status, differing occasionally in the details of how this should happen. At times parties suggest that the constitution should be amended to make Latvia a bilingual nation, speaking both Latvian and Russian. Other parties have suggested that Russian should be an official language of any municipality where the Russian population is greater than twenty percent. Many parties explicitly mention the school system as an important venue for language rights, arguing that Russian language education should be available at primary and university levels. For Human Rights in United Latvia also mentions the courts as a venue where officials should be required to provide services in Russian. Less often than language, some parties also mention specific policy areas where Russians may have preferences that diverge from the Latvian majority. For Human Rights in a United Latvia, for example, argued that providing support for small and medium-sized business was an important part of ensuring justice for ethnic Russians. Most Russian families had immigrated to cities and worked disproportionately as shopkeepers and other kinds of entrepreneurs, and therefore could benefit from subsidies and tax breaks to small business owners. Small business support is mentioned in platforms across the ethnic divide, with even ethnic Latvian parties like the National Alliance proposing similar tax incentives and subsidies. Yet for Human Rights in

187 a United Latvia specifically makes this an ethnic issue, saying: “Particular attention should be paid to the needs of small and medium-sized businesses, which are the economic basis of the Russian-speaking community.” Other policy areas affecting ethnic Russians stem from their status as recent migrants. Given that the bulk of the Russian population in Latvia either emigrated from what is now the Russian federation, or is descended from someone who did, they are far more likely than ethnic Latvians to have family members who live abroad. On this basis the Latvian Russian Citizens’ Party argued that visa policy was an ethnic issue, as onerous travel document requirements would prevent ethnic Russians from visiting loved ones. Some parties also explicitly announce their support for benefits for non-citizens, arguing that any permanent resident (including those without citizens’ passports) should be entitled to vote in local elections, or that exemptions from language examinations should be given to naturalization applicants over a certain age. The final way that ethnic Russian parties signal their ethnic affiliation is by emphasizing the historical contributions of ethnic Russians in Latvia. Given the role that historical commemoration had played in the independence movements, and the way that ethnic Latvian parties rely on pre-Soviet history to demonstrate their ethnic identities, many Russian parties specifically use historical references to Russia as a way of mobilizing supporters. For Human Rights in a United Latvia stated in its program that, “we respect history and Latvia’s heritage of cultural diversity. We are against neo-Nazism, and the historical revisionism of the Second World War.” The party went on to demand full benefits for World War II veterans, claiming that, “Veterans of the Anti-Hitler coalition must be given official status and social entitlements.” In this way, the party challenged the Latvian nationalist interpretation that the outcome of the war was the shameful subjugation of the Latvian people by presenting the defeat of fascism as an important Soviet achievement. This difference of interpretation over the historical significance of World War II between Russians and Latvians is used by other Russian parties to signal ethnic affiliation to voters, especially over state support for monuments and commemorations. The Latvian Rus- sian Citizen’s Party specifically proposed a program called “Cemetery of the Russian Soldiers” which would be funded to ensure that Soviet gravestones and historical monuments dedicated to ethnic Russians would be maintained and preserved. The party also proposed that citizens from

188 the CIS countries be granted visa-free entry if they intended to visit war graves or memorials. The Latvian Russian Party similarly announced their intentions to preserve historical monuments in Latvia which had been erected during the Soviet period. Some parties reject the notion of appealing to a specific ethnic group, and support a notion of a multicultural and multiethnic Latvia. The Latvian Socialist party in 1995 echoed much of the Soviet Marxist nationality policy in avowing that, “the main task of cultural policy is to preserve the cultural heritage of all the peoples of Latvia, and their traditional spiritual environment.” The Last Party—so named because of its anti-establishment positions against partisanship— campaigned on a platform of “respect for every Latvian in all our country’s diversity, regardless of nationality, language, gender, age, sexuality, origin, health, or other characteristics.”19 Unity echoed this sentiment, stating that its goal was to ensure equality before the law of all Latvians, “regardless of ethnicity, class, or sexual orientation.” The Progressive Centrist Party likewise stated that its primary goal was “Creating a socially secure, educated, cultured community and a strong middle class of its citizens, regardless of nationality, religion and social status.” Other parties show an embrace of Latvian diversity by drawing attention to the various historical, regional, and linguistic identities contained within Latvia. While for many ethnic Latvian parties, the “golden age” of Latvian history was the period of interwar statehood, other parties note that the Latvian ethnic and national identity itself hardly existed before that time, when Latvia was composed of separate linguistic groups. The People’s Control Party specifically endorsed revising the constitution to establish state support for the Latgalian dialect as a second, legally protected version of the Latvian language. Such a move would effectively make Latvia a bilingual country, and would have allowed dual-language street signs and governmental services in Latgale. The Union of Greens and Farmers, as well as For Human Rights in a United Latvia also supported some form of state sanctions for the Latgalian language, while the Social Justice Party drew attention to the social and ethnic diversity of Latvian society by announcing its support for development funds for Latgalian, Livonian, and Roma communities. Yet the most common way by far that parties avoid ethnic identification is simply never to mention identity politics at all, as if ethnic divides were not present in Latvia. A large number

19The party never managed to clear more than one percent of the vote, and was never seated in Parliament.

189 a b c

Figure 6.2: Latvian Mainstream Party Logos of parties campaign exclusively on economic issues, or good governance and anti-corruption, which would affect all ethnic groups. In 2014, Unity campaigned on a platform which not once mentioned issues of language, ethnicity, citizenship, or history. Its platform focused instead on tax policy, economic growth, European integration, healthcare, and development. When the party recommended developing cultural institutions in Latvia, they did not specifically mention which forms of art they would support, listing only “cultural infrastructure,” “cultural education,” and “folk art.” The party also listed regional development as a priority, but then explicitly mentioned all five distinct regions of Latvia, ensuring no group would be excluded. In another example, the People’s Control party in 2010 campaigned on a platform of returning power to “the people,” as well as reducing taxes, generous retirement benefits, and direct democracy, never once mentioning issues of nationality, language, or culture. In an ethnically divided country with a very recent history of ethnic mobilization and communal grievances, these parties which choose to make no mention at all of perhaps the most relevant social cleavage in the country are noticeably conspicuous in their silence.

6.3.3 A Note on Party Logos

Unlike in Bosnia, few parties in Latvia rely on party logos to convey their party aﬃliation. Figure 6.2 shows the logos of three of the top vote-getting parties in Latvia. In 6.2a, Harmony has chosen only a color and a font. The red is most certainly a nod to the social democratic

190 platform and its affiliation with other leftist parties throughout Europe, but has avoided any other distinguishing characteristics, including the fist, the rose, the handshake, the star, or any of the other graphic logos used by other socialist parties throughout the world. The same could be said of New Unity, whose logo is seen in 6.2b. The image of the braided cord suggests a coming together of various political entities but makes no reference to any specific ideology or identity. Figure 6.2c shows the logo of the Union of Greens and Farmers, whose logo conveys the party’s origins as an agrarian and ecological party. The fleche shape suggests a wheat spikelet, gold on one side and green on the other, to convey the union of two separate parties under one rural and environmentalist platform.20 Typical of most Latvian parties, none of these logos employ any overt ethnic identifiers. The problem for Latvian political entrepreneurs is that unlike in the former Yugoslavia, the most prominent national symbols of the Latvian nation-state are relatively new, and lack long- standing association with ethnic identities in the minds of voters. The national flag of Latvia and the Latvian Coat of Arms (Seen in figure 6.3) are both inventions of the interwar period, and thus reflect a relatively recent political entity, not a long-standing ethnic or cultural history. The carmine red color of the Latvian flag was chosen to represent the Latvian state based on apocryphal references in medieval German texts claiming that when Teutonic knights arrived in the Baltic region, the indigenous people who resisted them carried red banners (BNN 2012). The Latvian Coat of Arms, created in 1921, reflects the fact that Latvian nationalism was largely an attempt to unify peoples and regions who at the time considered themselves separate identity groups. The red lion on the left is the historic sigil of the Dukes of Kurzeme and Zemgale, while the silver griffin represents the Duchy of Vidzeme and Latgale. The seal is meant to represent the disparate political entities coming together in a single political unit (Latvijas Valsts Prezidenta Majas¯ Lapa 2017). The flag and the seal were seen as symbols of independent Latvia statehood—so much so that they banned by Soviet authorities, and the legalization of their display was an important demand of protesters in the lead-up to independence (Eglitis 2002,

20The shape may also be a nod to the Sign of Laima, the ancient Latvian goddess of destiny and nature. In traditional Latvian folk art, three downward pointing arrows represents an appeal to Laima for the blessings of nature, fertility, and luck. To my knowledge, the party has never directly conﬁrmed a link between the two, but former party leader and current Latvian president Raimonds Vejonis¯ has previously declared publicly that he considers himself a pagan (France 24 2015), so the link is not implausible.

191 a b

Figure 6.3: Latvian State Symbols: Established 1923 & 1921

Chapter 2). But both are more often seen as symbols of the Latvian state as a political entity, not an ethnic Latvian nation. Neither have particularly strong associations with an ethnic Latvian nationalism, having been ignored or banned for far longer than they were ever used in any official capacity. Neither are particularly helpful to political candidates in broadcasting a specific Latvian identity to potential supporters. There is, however, one set of exceptions to this in the far-right ethnic Latvian nationalist parties, which employ logos containing ancient Latvian pagan symbols. The most prominent example of this is the double diamond logo of the National Alliance in 6.4a. The National Alliance was formed in 2010 as a merger of two of the most prominent and outspoken Latvian nationalist parties: “All for Latvia!”, and “For Fatherland and Freedom/LNNK”—itself the product of an early merger of the Fatherland and Freedom Party, and the Latvian National Independence Movement. To convey the history behind the merger, the National Alliance includes not only the names of both of its constituent parties in the logo, but also shows an interlocking of both of the party’s original symbols. The logo bears a striking resemblance to a swastika, yet party members claim that the logo is not intended to be an endorsement of fascism. The swastika does have roots in the pagan Latvian religion, where it is usually referred to as the “Fire Cross” or the “Thunder Cross,” both linked to specific deities in the pre-Christian Latvian pantheon. The symbol is still a prominent feature in traditional Latvian embroidery and knitwear, sometimes

192 a b c

Figure 6.4: Ethnic Latvian & Far Right Parties confusing and offending international observers for whom the swastika pattern has a very different significance (Sheeter 2006). Despite its pre-Nazi origins in Latvia, the swastika remains most strongly associated with the Third Reich, and is designated a hate symbol by Latvian law. Along with the Soviet hammer and sickle, the use of a Nazi swastika is illegal in public protests and demonstrations, but the use of the “Thunder Cross” is not, creating a legal ambiguity that has often perplexed Latvian courts (Kamenska & Brands-Kehris 2008, p. 30). It hardly seems plausible, though, that the use of such overt fascist imagery was completely unintentional. Since no group but ethnic Latvians has ever followed Latvian pagan rites, the use of the Thunder Cross logo is clearly designed to make a strong statement on ethnic Latvian identification. Moreover, the very name “Thunder Cross” was also used by the paramilitary fascist organization that targeted Germans, Jews, Slavs, and other non-Latvians in the 1930’s, and which is widely seen as having collaborated with the Nazis in 1941 (Kasekamp 2006). In addition to courting far- right and anti-Semitic supporters, the use of the traditional Thunder Cross seems a purposeful and intentional way of showing dedication to an ethnic interpretation of Latvian nationalism. Party Leader Raivis Dzintars, himself the subject of controversy over anti-Semitic comments, has publicly defended the use of the Thunder Cross as a manifestation of indigenous Latvian traditions and national self-determination. Given the highly provocative and controversial appearance of the Thunder Cross, as well as the fact that it relies on pagan origins that are not only pre-Soviet but also pre-German, it is likely intended to convey an ethnic Latvian identity to voters. In short, with the noticeable exception of the far-right ethnic Latvian parties, politicians and

193 political parties in Latvian elections are far more likely to rely on information conveyed through party platforms than they are through visual depictions and logos. While party platforms do not shy away from explicit ethnic rhetoric, logos and visual depictions are for the most part relatively uninformative in the Latvian context.

6.4 Classifying Latvian Political Parties

In Chapter 4, I argued that political parties in Bosnia could be divided into a three-party typology based on how they use ethnic divisions in communicating with voters for the purpose of electoral mobilization. In this section, I argue that a similar typology is useful in understanding Latvian campaign mobilization. I ﬁrst outline that typology, and then address concerns about classifying parties as non-ethnic in the Latvian context

6.4.1 A Three-Part Typology

Several parties in Latvia clearly meet the deﬁnition of ethnic party. Nationalist parties like the National Alliance and Latvian Way promise that their policies will beneﬁt the Latvian people exclusively, and preserve the dominance of the Latvian ethnic group in the state. Parties like the Russian Party and the Latvian Russian Union are similarly ethnic in their promise to represent minority Russian interests in the Latvian parliament. But even parties more commonly accepted to be in the mainstream and the center of the Latvian political spectrum often appeal directly to coethnics as champions of ethnic interests. When Unity or the Union of Greens and Farmers promises to protect the power of “the nation,” they often make it clear that that nation is the ethnic Latvian community, whose language, culture, and communal interests will be protected at the expense of others. When parties argue that the ethnic community is under attack—either due to the community’s minority status in the case of Russians, or due to the historical and geopolitical vulnerability of the group in the case of Latvians—the party is implicitly claiming to rectify the situation by representing the group’s interests. Other parties should be seen anti-ethnic. Socialist parties especially often explicitly attack the “nationalist extremism,” or emphasize the historical diversity of Latvia and the greater Baltic

194 Type Characteristics Examples National Alliance, Claims to Represent Ethnic Ethnic Party Russian Party, New Community Era Rejects Ethnic Favoritism; Socialist Party of Anti-Ethnic Party Cultivates Multi-Ethnic Base Latvia, Dzimtene of Support Ignores Ethnic Division of Ethnically Ambigu- Country; Does not use Eth- Harmony Center ous Party nicity as Basis of Inclusion or (2014) Exclusion

Table 6.1: Latvian Party Typology area as a multi-national, multi-ethnic region. These parties promise not to represent any single group, but instead promise not to divide the country into ethnic groups, which they see as preventing the state from provide common goods like economic growth, good governance, or stability. Finally, a large number of parties meet the category of ethnically ambiguous by simply ignoring the ethnic divide in the country altogether. These parties promise policy beneﬁts that theoretically could be appealing to all members of society, but do not explicitly take a position on the ethnic divide within the country. These parties campaign on issues of class, ideology, region, or policy preference. These proposals may be supported disproportionately by one group over another, but the party does not actively claim to represent the interests of the group based on membership in any ethnic category.

6.4.2 Can a Latvian Party Really be “non-ethnic?”

It may be objected that a typology which sees certain parties as not employing ethnic divides is not applicable to the Latvian context. After all the ethnic divide is fundamental to many aspects Latvian politics, and one of the few persistent features of Latvia’s extremely volatile political party system. As in Bosnia, institutions have formalized ethnic divides into very practical matters of policy implementation. The citizenship laws hold one ethnic group as the “rightful” residents of the country, while others are the result of a decades-long criminal occupation. In such an environment, it is not unreasonable to think that no political act at all can be said to be

195 truly “non-ethnic.” Moreover, ethnic divisions have been salient to day-to-day life throughout Latvia’s entire history. The distinction between Latvian and Russian ethnic identities was at the core of the Soviet Union’s Russification policy during the twentieth century, and Latvian ethnic distinctiveness and historical heritage was an important mobilizing force against Soviet rule that lead to the independence of the state in the 1990’s. In the middle ages, Latvian ethnic identity came into existence as a result of feudal distinctions that held ethnic Latvians in a rural, agrarian position serving German nobility. It is not unreasonable to think that given the way ethnic identities have been so deeply embedded in nearly all relevant social distinctions in Latvia, that it is simply not possible for a party to avoid ethnic issues in any meaningful way. Such objections may stem from Latvia’s unique circumstances as a nationalizing state (Smith 1996, Agarin 2011). In this analysis, Latvia’s fundamental political conflict in the post-Soviet period is resolving the “national question,” in which the relationship between the political and cultural community is still ill-defined. According to Brubaker (1995), in a nationalizing state the key to political development and conflict is the so-called “triadic nexus,” between titular majorities, national minorities, and the external homelands to which the national minorities belong. The nationalizing question is fundamental to the very nature of the state and of society, and involves not just the setting of state policies, but the very reformulation of the ethno-national groups and of the state itself. In this sense, truly democratic politics cannot be discussed in Latvia because the question of to which group of “the people” the state is accountable has not yet been fully settled (Rustow 1970, Linz & Stepan 1996). This is a reasonable objection. Latvian citizenship policy over the past two decades has imposed certain linguistic and cultural standards as prerequisites to political participation. As a result, it is impossible to ignore the ethnic divide in the country, since ethnic identities and diversity must be addressed by every individual prior to political participation. Moreover, Latvia’s nationalizing state policies also seem to be changing the very nature of the “Russian” identity itself, reforming the way Latvian Russians see themselves and the relationship between their ethnic identity and the society in which they live (Laitin 1998). I do not want to dismiss the importance of these group-level disputes, or downplay the importance of these fundamental issues. But it is also important to realize that Latvia is still

196 an electoral democracy. Through its consolidation of electoral institutions, and its entry into NATO and the European Union, Latvia has committed to governance through electoral means. Latvia’s nationalizing state should therefore be seen as both an outcome and a context of democratic competition. Over the long term, it is an outcome: Latvia’s policies relating to citizenship, language, education, and media are those which have been enacted by democratically-elected politicians, chosen in free and fair elections by Latvian voters. While it is true that historical and structural factors may have predisposed Latvia towards nationalizing policies, it cannot be said that Latvian voters have not had opportunities to revisit these policies. As the discussion of campaign materials above suggests, maintaining or challenging those nationalizing policies has been an important point of contestation in nearly every election since independence from the USSR. These policies are maintained or altered based on the priorities of the leaders who get elected, which is an interactive product of the leaders who choose when and how to build political support, and the voters who decide who to put into office. While Latvia’s status as a nationalizing state may very well be one of the most important factors in understanding its political system, it is inaccurate to say that this context controls all aspects of its politics—it is not the case that Latvia is nationalizing and therefore elections nationalists, rather it elects nationalists, so it stays nationalizing. Moreover, the question of what exactly it means to be Latvian is not exactly settled in a way that has been unanimously accepted. As stated above, the Latvian “national” identity is itself a modern product: a linguistic and cultural amalgamation of what had been distinct regional identities whose dialects were mutually intelligible and who were all understood to be different from Russians and Germans. That identity is still somewhat contested today, as seen by the political entrepreneurs who debate the status of the Latgalian and Livonian languages. Even within the ethnic Latvian parties who claim to speak for their coethnics and exclude other ethnic groups, there is some debate over what traits define the identity. These parties argue over whether Latvian identity is fundamentally religious or secular, and even beyond that, whether it is historically Christian, or pagan. They offer competing views of Latvian identity which may be rural and traditional or urban and cosmopolitan. These questions are partially settled through electoral contests, and it is clear that candidates adopt a wide variety of positions and arguments

197 on this topic. Ignoring this variation as a result of prior assumptions risks missing key insights into politics not just in Latvia, but in all ethnically diverse societies. Finally, the framework which allows for the possibility of parties to be non-ethnic more adequately reﬂects the actual agency that voters and political elites have in making political decisions. In the Latvian case especially, where all parties can propose 4,000 characters of whatever positions and claims that they want to make, leaders and candidates have a lot of leeway in what they want to say to voters. By looking at the entire range of proposals and adequately cataloging their diversity, it is possible to observe not only the ultimate outcome of elections, but the outcomes that were not chosen by voters. Many parties choose not to campaign on ethnic issues, at times even attacking the nationalizing policies of the state outright. That in itself is important: it suggests that some leaders believe that those policies are not in the best interest of the Latvian state—or, at the very least, that attacking those policies could be a viable path to electoral victory. But acknowledging that those arguments are made also helps understand the range of options available to voters and understanding why they make the choices that maintain ethnic divides as important political cleavages. By separating political decisions from identity characteristics, we avoid essentializing ethnic identities, and can better understand the circumstances under which ethnic identities inﬂuence voting decisions.

6.5 Conclusion

Unlike in Bosnia, the ethnic diversity in Latvia is a much more recent development, largely the result of Soviet-era migration patterns. Furthermore, the divide between the two ethnic groups is also much wider than in the Yugoslav case due to differences in history, language, and culture. Latvians speak a Baltic language, and have historically been politically dominated by the much larger powers surrounding them. Russians speak a Slavic language, with a different alphabet and virtually no intelligibility with Latvian, and are for the most part present in Latvia as a result of the Russification policies of the Soviet Union. Following democratization and independence from the USSR in the 1990’s, this ethnic divide become an important political cleavage in voting behavior. Both ethnic Latvian and ethnic Russian candidates have campaigned for support from

198 coethnics throughout the post-Soviet period, often relying on differences in culture, language, and differences in the historical heritage between the two groups to signal to voters their ethnic affiliation. Yet during this period there is also a group of political leaders who do not campaign on ethnic issues, choosing instead to mobilize voters by advocating for a multi-ethnic civic nationalism, or by ignoring ethnic and identity divides to focus on economic issues. Just as in Bosnia, Latvian voters therefore find themselves in an ethnically divided society with a wide range of parties to choose from, which employ ethnic identities in different ways. But unlike in Bosnia, there is virtually no institutional protection for ethnic minority groups, as Latvia has restored its electoral institutions almost exactly as they existed prior to the arrival of the Russian minority in the twentieth century. In Chapter 2, I argued that majority and minority group members had divergent incentives to support ethnic parties. In Chapter 5, I showed that that framework was helpful in identifying voter-level tendencies to split their votes between ethnic and non-ethnic parties. In the next chapter, I employ a dataset based on the classification of Latvia campaign rhetoric presented in this chapter to analyze the over-time variation in ethnic party success in Latvia.

199 7

Latvian Elections over Time

In the previous chapter, I outlined the characteristics of the ethnic divide in Latvia, how Latvian political entrepreneurs use ethnic identities in mobilizing voters. I now turn to an analysis of the success of these mobilization strategies. These proposals are not all equally persuasive to voters, and identifying which types of parties are most likely to succeed will help to explain the role of ethnic identities in electoral politics. In this chapter, I show that ethnic majority Latvians and ethnic minority Russians exhibit divergent patterns of ethnic voting in the democratic period following the collapse of the Soviet Union. Since the mid-1990’s, appeals to ethnic Latvian identity have increased. Candidates who previously campaigned on exclusively non-ethnic platforms show a predisposition to adopt positions protecting an ethnically-deﬁned Latvian culture, legal protections for Latvian-language speakers, and a rejection of non-Latvian voters. Explicit Russian identiﬁcation, on the other hand, has declined, and candidates who present themselves as champions of Russian interests show an over-time tendency to abandon explicit ethnic protections. These candidates instead advocate positions ostensibly acceptable to members of all ethnic groups, such as civic nationalism, multiculturalism, economic redistribution and populism. As a result, the ethnic cleavage in Latvian politics should not be understood as “ethnic Latvian identity versus ethnic Russian identity”, as it is often presented, but instead “ethnic Latvian identity versus non-ethnic identity.” In Latvia, ethnic campaigning has become a privilege of the majority group only. This chapter will show that during Latvia’s period of democratic consolidation, ethnic voting

200 has remained relatively consistent for ethnic Latvian parties as a share of the total vote, but has declined for ethnic Russian parties. As votes going to ethnic Russian parties have decreased, the share going to parties espousing multi-culturalism or non-ethnic identities has increased, as ethnic Russian voters have shifted over time from supporting ethnic parties to supporting non- ethnic parties. This trend is largely the result of political entrepreneurs changing the platforms on which they campaign. As Latvian democracy consolidates, candidates who had previously espoused an ethnic Russian platform are more likely to change their positions from ethnic to non- ethnic. Candidates who had previously campaigned on a non-ethnic platform are more likely to change to an ethnic Latvian one, but very few candidates adopt a Russian ethnic platform after having previously taken a non-ethnic position. Both ethnic Latvians and ethnic Russians have reasons to support ethnic group representation, since both groups are large enough to be potentially lucrative targets of expropriation by the other group. But as ethnic Latvians are an absolute majority of the country, and ethnic Russians are not, the likelihood of winning elections and wielding power is much higher for the Latvians than the Russians. This chapter proceeds as follows: in the next section, I discuss why analyzing over-time variation in Latvian party politics is an appropriate way to test a theory predicting divergent incentives between majority and minority group voters. I then describe the data set—derived from the observations of the previous chapter—which I use to classify ethnic appeals to voters in Latvia and measure their relative success in attracting voter support. In section three, I present a quantitative analysis of the success of these appeals over time, and measure the likelihood of candidates making each type of appeal. In the fourth section, I show what those patterns look like in actual Latvian candidates and give examples of the way that candidates rebrand and sort themselves in to ethnic and non-ethnic parties over time. The ﬁnal section concludes.

7.1 Over-Time Variation in Latvia

This chapter primarily looks at variation in Latvian electoral outcomes over time. This is based on the premise that parties, candidates, and voters are likely to function diﬀerently in Latvia immediately after independence than they are after several years of democratic governance. The

201 years immediately following democratization are often the most chaotic and unstable in any country. Institutions are still weak, and politicians often lack the experience and institutional resources that political actors possess in long-established democracies. But through a processes of political learning and Darwinian competition, political actors adjust their strategies. Can- didates who are unable to attract voter support or adapt to more efficient electoral strategies lose or drop out, and with time the party system consolidates around those parties and actors which are most electorally viable. The argument of this dissertation is that ethnic voting is a reliable strategy only for groups that have high benefits of ethnic representation, and relative ease of winning office by making ethnic appeals to voters. In the case of Latvia, this theory suggests that ethnic majority Latvian appeals are more likely to be successful than appeals to ethnic minority Russians. Over time, these divergent incentives shape the characteristics of winning candidates during the period of democratic consolidation. Several scholars have noted the importance of time as an explanatory variable immediately following democratization in shaping political party systems. In a cross-national survey of party system development, Peter Mair observed that “while cleavage structures act to stabilize electorates, they tend to do so slowly, with the result that almost regardless of region or timing, the early years of party systems, and the early years of newly democratized electorates, tend to prove the most unstable (Mair 1998, p.182).” This instability and rapid change is especially pronounced in the post-communist world. As opposed to first-wave countries in Western Europe, post-communist countries like Latvia democratized in an environment where civil society organizations had been actively suppressed by the state, and in many areas largely eradicated. While the Latvian independence movement was led by the few civil society organizations that existed in Latvia at the time (e.g., environmental organizations, cultural preservation societies, and the Popular Front), these groups hardly had the mass-membership or all-encompassing support as the trade unions or churches which were key actors in first- and second-wave democratizations. Political parties in Latvia for the most part had to be formed or revived from non-existence, and the link between social identities and political identities was relatively weak. In this somewhat tumultuous and uncertain environment, the choices made in the day-to-day business of governance may ultimately shape the resulting regime, serving to consolidate both democ-

202 racy as a regime type, and the political party system which structures future electoral contests (Morlino 1998). In the absence of pre-existing and well-established party institutions, political actors are forced to legislate, campaign, and interact with voters in a way that responds directly to the immediate practical challenges of governance and reelection. Here, where new actors emerge with relative ease and the balance between existing actors has been completely upset, outcomes are likely to be less stable. Until actors learn what constitutes a “good” decision—or at least are voted out of office for making “bad” ones— the range of viable political players and possible policy outcomes is likely to be much wider than in more institutionalized long-standing states. This temporary instability has been documented empirically among elites and voters. One of the clearest signs is the tendency of recently democratized countries to see an explosion of new political parties immediately following democratization, only to taper off after successive rounds of political contestation. In the absence of strong civil society organizations or political parties with established rules for succession of leadership, there is a high degree of uncertainty over who constitutes a viable candidate, or which ideology is popular. This uncertainty is often not resolved until actual elections provide new information on which positions are viable, and which are not (Rose & Munro 2003). Further complicating things is the fact that the most pressing political issues facing the country—and therefore those issues which are likely to be the most important for soliciting voters support—are likely to change over time as the country transitions from a command economy to a market economy. Kitschelt (1992) argues that in the immediate post-communist period, foundational questions about regime type and economic transition are likely to dominate elections, which may favor one set of actors or policies over another. Once those issues are settled, other policies and social cleavages are likely to become more relevant, causing a change over time in the strategies elites use to build linkages with voters. How the deposed communist party and its candidates transitions from the only legal political entity to one of several electoral competitors can also have drastic impact on not only its own electoral fortunes, but also the strategies of its competitors (Sakwa 1998, Grzymala-Busse 2002). Voters also go through a period of political learning immediately following democratization. Evans & Whitefield (1998) finds that post-Soviet voters readjust of their own ideology follow-

203 ing the introduction of free elections, which they attributes to the reconﬁguration of political competition following the collapse of communism. Oﬃcial Soviet ideology had held that the “left” was the existing regime, while the “right” consisted of reactionary anti-communists. Fol- lowing the demise of the Soviet Union, liberalizing, pro-reformist parties branded themselves as the left, while supporters of the Soviet regime were considered to be on the right, suggesting that the borders of the policy space were being reformed by political entrepreneurs as a result of the transition. Brader & Tucker (2001) note that what could be considered partisanship takes several years to develop in any meaningful sense in post-communist Russia. This, they argue is because partisan and ideological labels only take hold if they show consistent usefulness in line with a voter’s personal experience. These are time-dependent processes, and as such strong partisan or ideological identities can emerge only years after the democratic regime has been installed. These observations are consistent with expectations from the literature on partisanship in more longstanding democracies, which suggest that political identities are established over the long term, reinforced by socialization, and personal experience (Converse 1969). Unlike Bosnia, Latvia is a unitary system. There are no autonomous local governing entities, and there are no explicit protections for ethnic minorities. As there is only one level of government that sets the most relevant policies, Latvian citizens are either always a majority or always a minority. Although ethnic Latvians do constitute a majority of the country, the size of the ethnic Russian community is very large. Policies which protect the interests of ethnic Latvians only would therefore exclude a very large number of people. Redistributing resources away from ethnic Russians to ethnic Latvians is potentially an extremely lucrative opportunity, since the two groups are near parity in size. Moreover, the argument that ethnic Latvians could be oppressed by minority Russians does not likely seem implausible to Latvian voters, as many are old enough to remember the Soviet period when such oppression was quite real. Latvians therefore have strong incentives to desire ethnic representation. Likewise, ethnic Russians also have strong reasons to desire group representation in parliament. As a minority group in a nationalizing state, they are the most likely to see their language diminish in importance. Given the centralized nature of the Latvian state, and the importance of historical commemoration to the independence movement, educational curricula are likely to focus on ethnic Latvian inter-

204 pretations of history that may be insensitive or offensive to ethnic Russians. The two groups diverge, though, in the likelihood of actually obtaining power on the basis of ethnic mobilization. By campaigning to represent the interests of ethnic Latvians over other ethnicities, a party limits its appeal to roughly 60% of the voting population. By appealing to Russian speakers, a party limits itself to 40% of the vote. This means that ethnic Russian candidates pay a steeper price when appealing to coethnics, and impose a greater limit on their support than an ethnic Latvian candidate does pursuing the same strategy. In other words, under perfect ethnic census voting—where all Latvians voted for one party and all Russians voted for another—ethnic Russians would be permanently excluded from power. Because ethnic Latvians constitute an absolute majority of voters, ethnic mobilization ultimately benefits Latvians more than Russians. Both groups have strong reasons to want the benefits that could come from ethnic representation in policy making. But the ethnic Latvian community has a much higher ease of access to actually winning power than the ethnic Russians do. Since ethnic Latvians are high on both dimensions, while the Russians are high only on one of them, ethnic voting is most likely among Latvians not Russians. Over time, the process of political learning should favor Latvian candidates who make ethnic appeals, but not Russians, setting the two sides on different trajectories. Ethnic Latvian candidates who appeal to voters on the basis of ethnicity have an advantage over those that don’t, whereas ethnic Russian candidates who don’t appeal to voters on the basis of ethnicity have an advantage over candidates who do. Through the process of political learning during democratic consolidation, Latvian candidates should therefore “learn” to continue or increase their ethnic messaging, whereas Russian candidates should “learn” to avoid it. This produces several testable hypotheses of Latvian party system consolidation. Successful ethnic appeals made by candidate to ethnic Latvians should increase or remain high over time, since those voters have both much to gain through ethnic representation, and a great likelihood of winning. Ethnic appeals to Russians, on the other hand, should decrease or remain low over time, since Russians are less likely to be able ultimately to enact policies. Since a candidate who makes appeals to ethnic Latvians is likely to find those appeals rewarded at the polls, those candidates are likely to continue to follow that strategy. Candidates who make appeals to ethnic

205 7.1 Ethnic appeals to Latvians should increase or remain high over time. 7.2 Ethnic appeals to Russians should decrease or remain low over time. 7.3 Candidates who make appeals to ethnic Latvians will continue to do so. Candidates who make appeals to ethnic Russians will change their ap- 7.4 peals. Candidates who make non-ethnic appeals will be more likely to change 7.5 to ethnic Latvian appeals than ethnic Russian appeals.

Table 7.1: Hypotheses

Latvians will therefore continue to do so over time. A candidate who makes appeals to ethnic Russians, however, is likely to ﬁnd those appeals ultimately futile, so those candidates will likely change strategies to make non-ethnic appeals. Any candidate who starts by making non-ethnic appeals has no incentive to change to supporting an ethnic minority position, but may see reasons to support the ethnic majority group. Those candidates will therefore be more likely to change to ethnic Latvian appeals than ethnic Russian appeals. A list of these hypotheses is presented in Table 7.1

7.2 The Data Set

One of the greatest challenges to measuring over-time variation in ethnic mobilization in new democracies is finding valid and reliable measures. “How ethnic” a party should be considered is an extremely difficult concept to measure, given that ethnic identification is often communicated subtly. Existing measurement strategies are especially ill-suited to capturing over-time variation, largely because they rely on indicators which change extremely slowly, or do not match the information available to voters at the time they made their voting decision. Expert surveys, one of the most common methods of measuring a party’s ethnic positioning, rely on a subjective assessment of a party’s positions which show little variation in the short-term. Consciously and subconsciously, experts make their assessments on a rolling basis, synthesizing observations over the medium-and long-term into a single estimation of the party’s identity. Such measurement is ill-suited to assess the moment that a party stopped or started “being ethnic.” Likewise, roll-call vote tallies—another system used to measure a party’s positioning—suffer from problems in measuring over-time variation. This study seeks to address ethnic identification and

206 electoral success, and as such must rely on information available to voters at the time of the election in which they voted. Roll-call votes measured after an election would therfore not be appropriate, since there is no guarantee that these votes were indicative of the way the party campaigned before they were elected. These problems are especially severe in the Latvian context. As discussed in the previous chapter, Latvia has an extremely volatile party system. Latvian political parties frequently form de novo, or consolidate as the merger of multiple parties to create a new party. Sometimes these are relatively straight forward, such as when the “Fatherland and Freedom” party merged with the “Latvian National Independence Movement” to become the “Alliance of Fatherland and Freedom and Latvian National Independence Movement.” Others are less obvious, such as when the “Latvian Popular Front” changed its name to the “Christian People’s Party.” The name change suggests a rebranding from a nationalist party to a religious one, yet the party’s leadership remained mostly the same. In fact, of the 118 parties in this study, 90 of them contested only a single election. Measuring parties’ identities and positions over time presents a serious practical challenge. Since candidates frequently change affiliation as parties reform, merge, and dissolve, the standards for assessing continuity are unclear. To resolve this issue, I return to the Latvian pre-electoral party programs. As discussed in the previous chapter, every political candidate in Latvia is required to appear on a party list, and every party list must be submitted with a party program. Since no candidate can appear on a ballot without signing her name to a platform, it is possible to assign a specific platform to literally every candidate who has ever run for national office in post-Soviet Latvia. This includes both winning candidates and losing candidates, resulting in no data loss from candidates deemed “insignificant” by other measurement strategies. I first obtained party platforms from all elections since 1995 directly from the Latvian Elec- tion Commission in Riga. I then hand-coded a total of 118 programs over all electoral cycles from 1995 to 2014 according to pre-determined coding criteria. The standard imposed is that outline in Chapter 2, and used in Chapter 5 to code Bosnian political parties: to be considered ethnic, the party must clearly identify as representing a specific ethnic group. It must also exclude other ethnic groups from their agenda, so that the voter can infer which groups the party

207 with not seek to support once in office. In short, if a reasonably intelligent voter would identify which ethnic group a party claimed to support using the information in the party program, I code it as ethnic. In practice, this means that those parties which were inductively described as anti-ethnic or ethnically ambiguous in the previous chapter are collapsed into the same category. These data were then mapped to the full candidate lists and vote totals for all candidates registered to contest national-level parliamentary elections. Unfortunately, candidate lists from the 1995 election are not publicly available, and so while party programs from that election are coded, candidate-level data from that election is not included in the dataset. Nevertheless, the resulting data set does include 8, 767 observations listing every candidate who contested elections in the period surveyed, and whether or not they explicitly identified with a specific ethnic group in each electoral contest. The standard does result in some potentially surprising coding decisions the most controversial of which is the classification of Harmony in some years as a non-ethnic party. In several of the elections under study, Harmony completely ignores issues of ethnicity in its party platforms, choosing instead to campaign on economic issues. It is therefore classified as non-ethnic in those years. But Harmony is supported overwhelmingly by the Russian population in Latvia. Many of its candidates are themselves ethnic Russians, and many journalistic sources describe it as a “Russian party.” It is not inaccurate to describe Harmony as a Russian party, but it is inappropriate for the purposes of this study. Ethnic Russians cannot choose to stop being Russian. However, when Russians run for office they can choose whether or not to make their identities central to their political brand. When Russians vote, they can also choose whether to support a party that claims to represent their group interests. These decisions are the subject of this study. In order to understand the influence of ethnic identities in political decision-making, it is crucial to have measures of those decisions that are not based on assumptions about the relationship to identity. Classifying Harmony is a Russian party because most of its voters and candidates are Russian is therefore inappropriate. Moreover, the measurement is consistent with the way Harmony chooses to present itself to the world. Harmony’s leadership goes out of its way to distance itself from its Russian origins. It has conscientiously recruited ethnic Latvians to join its candidate lists, so much so that leadership of the party now has only one ethnic Russian

208 (Khlakovskiy 2018). The party has also publicly renounced its ties to United Russia, Putin’s governing party in the Russian Federation, choosing instead to join the Party of European So- cialists which is composed of western left-of center parties like the British Labour Party, the French Socialist Party, and the German Social Democratic Party (Bershidsky 2018). This non- ethnic determination is also consistent with the comments made by party leadership in the past which seem to seem to antagonize the party’s ethnic Russian base, such as when Nils Ušakovs, the former mayor of Riga and Harmony’s leader, declared that, “it’s better to take the citizenship exam than to sit around and complain.” Ušakovs noted that he, himself, was a non-citizen in the 1990’s, and that naturalization is not overly difficult because, “it’s not a very hard test (DELFI.LV 2012).” It may be of concern that party platforms are little more than empty talk. Parties can say whatever they want in the program, and accordingly these documents may not be good indicators of reality. In some cases, that is almost certainly true. Several parties submitted platforms that strain credibility with promises to guarantee enormous social benefits miraculously financed by eliminating all taxes. Yet in this study, these platforms are not taken as an indication of what a party will actually do if elected. Rather, they are meant to indicate how a party wishes to present itself to voters. If a party uses some of its valuable 4,000 character allotment to declare that it will lower taxes and raise incomes, we cannot infer that it will actually be able to do (or that it even really intends to do) any of this, as actual policy outcomes will depend on the party’s true intentions, post-election coalition bargaining, and the mathematical constraints of public finance. What we can infer is that this party wants to appeal to everyone who pays taxes and relies on income. This type of inference is especially useful if a party explicitly identifies itself as a representative of a specific ethnic group, a champion of a particular ethnic interests, or hostile towards a set of ethnic outsiders. From these statements, we can infer that the party wants to be perceived as belonging to a specific ethnic group. In this way, the measurement strategy captures the variation most relevant to the research question.

209 7.3 The General Trajectory of Latvian Party Consolidation

Over time, voter support for ethnic Latvian parties has remained relatively constant, while the support for ethnic Russian parties and multi-ethnic parties has varied. Over time, fewer and fewer voters support candidates making overt appeals to Russian ethnic identities, while more and more voters are supporting candidates making multi-ethnic appeals or ignoring identity all together. Moreover, candidates themselves respond to incentives to adopt or abandon ethnic messaging based on majority or minority status. I show this by producing a series of estimated probabilities of a variety of candidate outcomes and behaviors derived from multinomial logit models. Specifically, in this section I show that over time the candidates who make Russian ethnic appeals are less likely to win elections at all, while candidates who make non-ethnic appeals or ethnic Latvian appeals increase their relative likelihood of victory. I also show that the type of appeal made to voters has substantial impact on the likelihood of changing the nature of one’s platform between electoral cycles. Candidates who changed the nature of their ethnic affiliation between electoral cycles—either by revising their party platform or by changing parties—are far more likely to adopt either a non-ethnic identity, or an ethnic Latvian identity. Virtually no candidate ever abandons a non- ethnic platform to adopt a Russian one. These trends are further evident with new candidates. Candidates who contest elections for the first time over the period of study are far more likely to espouse an ethnic Latvian platform or a multiethnic platform than an ethnic Russian platform. In the end, ethnic Russian interests are more likely to be subsumed into a universalist policy of multiculturalism or multi-ethnic populism, while ethnic Latvian interests can be much more explicitly used as a tool by politicians to win elections.

7.3.1 Party Vote Share over Time

In Figure 7.1, I show the percentage of the vote going to parties that make each of three types of appeals: ethnic Latvian, ethnic Russian, and non-ethnic. Appeals to ethnic Latvians remain relatively constant over the entire period. With the one notable exception of the election in 2002, parties portraying themselves as champions of ethnic Latvian interests win roughly 50%

210 Latvian Ethnic Identiﬁcation 1.00

0.75

0.50

0.25

0.00 1995 1998 2002 2006 2010 2011 2014 Year Russian Ethnic Identiﬁcation 1.00

0.75

0.50

0.25

0.00 1995 1998 2002 2006 2010 2011 2014 Year Non-Ethnic Identiﬁcation 1.00

0.75

0.50

0.25

0.00 1995 1998 2002 2006 2010 2011 2014 Year

Figure 7.1: Vote Share by Electoral Appeals in Latvia: 1995-2014 211 of the votes in all Latvian elections. The drop in 2002 may be the result of Latvia’s referendum to join the EU in 2003. Many parties that support the EU campaigned that year by describing themselves as internationalists or European integrationists, rather than in ethnic terms. Some of those that were against Latvia joining the EU did so by describing themselves as committed to national sovereignty and the independence of the Latvian state, which technically they defended on non-ethnically deﬁned terms. Many of the rightist and Latvian nationalist parties therefore, for this election only, did not make explicitly ethnically deﬁned appeals.1 However, the general pattern returned in 2006, providing strong support for 7.1.

Consistent with 7.2, ethnic Russian appeals follow a different trajectory. Parties that make explicit appeals to Russian ethnic interests never reach the level of ethnic Latvian appeals. This most likely reflects the natural ceiling imposed on ethnic Russian appeals by demographics. Since ethnic Russians make up less than 40% of the population of Latvia, it is extremely unlikely that ethnic Russian parties would ever amass more than 40% of the vote. The highest value this ever reaches is 22.7% in 2006. Yet the over-time variation is quite remarkable. The gradual increase of each election cycle is readily apparent, but drops off suddenly following the 2006 election. While the gradual increase in the first four elections is most likely explainable by the increasing number of Russian voters gaining citizenship throughout this period, citizenship laws fail to account for the sudden drop off in 2010. Ironically, the greatest number of naturalizations happened in the mid-2000’s (see Figure 6.1), and by 2010 the number of ethnic Russians eligible to vote in Parliamentary elections was almost certainly higher than it was in 2006, yet the explicit Russian ethnic identification plays a relatively minor role in the Latvian party system beyond this point. Instead, Russians voters tend to support parties not campaigning on specific ethnic identification, but rather on issues that appeal to ethnic minority voters on non-ethnic grounds. This is reflected in the corresponding increase in the support for non-ethnic parties. Figure 7.1 depicts a party system-level overview of Latvian politics. The figure suggests that parties which campaign on explicit Russian ethnic identification wane in importance over

1See Mikkel & Pridham (2004), Ikstens (2007), and Schulze (2018). Euroscepticism was much lower in Latvia in comparison to other Baltic states, possibly because it was presented in terms that comported with nationalist ideology and ethnic Latvian interests. Both pro- and anti-European Union candidates defended their position with relation to Latvian “sovereignty,” seeing threats to Latvian independence coming from Russia or the EU respectively. However, in the platforms, many of these parties defend these positions with relation to the Latvian state, not the ethnically-deﬁned Latvian nation.

212 time, as a smaller and smaller segment of Latvian voters support them. Parties which make ethnic Latvian appeals, though, do not diminish in the same way, maintaining constant support. These observations are focused on voting outcomes, and combine two separate processes: the politicians proposing positions, and then voters choosing what to support. I now turn to examine these dynamics from the actors’ perspectives through a separate quantitative analyses of candidates.

7.3.2 Candidate-Level Analysis

The analyses that follow use the candidate-year as the unit of analysis, which allows for an individual to appear in the data set multiple times. If, for example, a candidate appeared on a party list at every election between 1998 and 2014, then that candidate appears in the dataset six separate times. Indexing by year has the beneﬁt that candidates are not assumed to be consistent in their ethnic messaging. A candidate can campaign on ethnic appeals in one election, but not in the next, (or vice-versa) and that variation will be captured in this measurement strategy. This approach produces a very large data set with a high level of statistical power. Over six electoral cycles, a total of 8, 767 observations are recorded, produced by 4, 508 unique individuals.2 Of those 4, 508, a total of 1, 041 individuals contested multiple elections, creating a complete picture of the Latvian political class and all possible legislators of the country. These large numbers facilitate analyses using even relatively demanding statistical models requiring large-

= data sets. The use of candidate-years also takes full advantage of the Latvian legal environment with regards to party campaign messaging. Every candidate who has ever run for parliament in Latvia has been required to appear on a party list, and to sign their name to an oﬃcial party platform which is distributed free of charge to voters. Every electoral cycle, all candidates must publicly declare their policy positions, and must revise those positions for every election. It is therefore possible not only to compare every candidate to every other candidate, but also every candidate’s current positions to her previous positions. Since the classiﬁcation outlined in the previous chapter rests on an exhaustive and mutually exclusive typology drawing from

2The 4, 508 ﬁgure and the analyses which follow are based on the assumption that individuals who have the same name, spelled the same way are the same person.

213 party platforms, and because every candidate has signed her name to a party platform at every election she has contested, the data set contains 100% coverage of every candidate for parliament in every election under study. Nevertheless, the linking of candidate-years to party platforms in order to provide measure- ments on candidate ethnic positioning is not without some controversy. The biggest concern is that because party lists consist of large number of candidates, the number of candidates is much larger than the number of party platforms. The project thus uses 99 party documents to code 8, 767 individuals. It may be argued that since party platforms are generated collectively by multiple candidates and party members, that a candidate-level analysis is inappropriate, and that a party-level analysis would be more suitable. This concerns has some validity, but should be taken as cause to limit the appropriate conclusions that can be drawn from the analyses that follow, not to reject them entirely. It is true that since the candidate coding schemes are drawn from sources which do not produce variation within parties, the analyses cannot provide any information on which candidate within a party list is most likely to eventually be seated in parliament. This is unfortunate, given how significant intra-party competition is in Latvia. But the most important variation for this study—the difference between candidates who make ethnic appeals and those that do not—exists at the party level not the individual level. This variation is captured by the coding of party platforms, and therefore the measurement strategy is still valid for the research question. Moreover, the use of candidate-years is the easiest way to avoid difficult, almost existential questions about what constitutes a “political party” in Latvia which could undermine any attempt at systematic study. Given the extremely high volatility, and the frequent occurrence of schisms, mergers, and new party foundations, over-time analyses of parties positioning are extremely difficult. If a party has a completely new name, new logo, and new platform, but almost all of its candidates had previously been members of the same party, is it a new party? Given that parties are frequently formed and registered as combinations of multiple loosely-organized factions, factional shifting can create big difficulties for defining party continuity. It is not uncommon for a party to fold and one portion of its candidates go into one party, while the rest go into another. Party-level measurement ignores those dynamics, and require potentially controversial a priori

214 decisions about which parties are continuations of previous parties, and which are new. Using the candidate-year as the unit of analysis, I conduct a regression analysis using a logit model with the dependent variable as a dichotomous indicator of whether the candidate won a seat in parliament, and the independent variable a vector of dummy variables indicating whether the candidate explicitly identified as either a champion of ethnic Latvian interests, ethnic Russian interests, or neither. In order to avoid collinearity, the coding scheme treats non-ethnic appeals as a residual category. The strength of the logit analysis here is that it is fundamentally probabilistic. The theory does not presume a deterministic relationship between ethnic campaigning and electoral appeals: it merely predicts that there should be differences in the success rates of ethnic appeals between majority and minority group. I therefore use the estimates produced by the analysis to generate predicted probabilities with 95% confidence intervals for each type of appeal at each electoral cycle, which I present in Figure 7.2. The G axis on this chart represents the elections, while the H axis represents the estimated probability that a candidate would win that election after making either an ethnic appeal to Latvians, an ethnic appeal to Russians, or no ethnic appeal at all. In 1998, for example, an estimated 6.1% of candidates who made ethnic appeals to Latvians won election to the national parliament, while an estimated 21% of candidates who made ethnic appeals to Russians won, and 2.9% of candidates who campaigned on non-ethnic platforms won. The predictions are somewhat noisy. This is not particularly surprising, as the measurement strategy puts a large amount of variation into the error term. The classification system measures variation only on ethnic positioning, but no party in this data set takes positions only on ethnic identification. Most platforms which explicitly declare for either ethnic Latvians or Russians also articulate positions on foreign policy, taxes, social spending, etc., which these data do not address. Those positions are nevertheless likely highly relevant in the decision-making process of Latvian voters. Many of the most salient issues in electoral politics, such as incumbents versus opposition and retrospective and prospective assessments of candidates, are ignored here. The “bouncing” pattern seen in some of the estimates is likely the result of government turnover, as voters support for specific parties ebbs and flows over time. However, there is a clear divergence in the long-term trajectories of the candidate appeals to ethnic Latvian and non-ethnic parties

215 Appeals to Ethnic Latvians 0.15

0.10

0.05 Probability of Election

0.00 1998 2002 2006 2010 2011 2014

Appeals to Ethnic Russians 0.3

0.2

0.1 Probability of Election

0.0 1998 2002 2006 2010 2011 2014

Non-Ethnic Appeals 0.15

0.10

0.05 Probability of Election

0.00 1998 2002 2006 2010 2011 2014

Figure 7.2: Probability of Election in Latvia by Appeal Type over Time 216 in contrast to ethnic Russian appeals. Consistent with 7.3 and 7.5, the success of both ethnic Latvian appeals and non-ethnic appeals increases in the period under study; while there is some variation year-to-year, there is a distinct upward trajectory for both. Russian appeal success, however, drops off, eventually reaching zero. This suggests that a candidate today making an appeal to ethnic Latvians has a higher chance of winning office as a result of that appeal than she did in 1998, and is consistent with 7.4. The same is true of a candidate making non-ethnic appeals, or ignoring the issue entirely. But the exact opposite is true for a candidate making an ethnic appeal to Russian voters. Whereas in 1998 a Russian ethnic appeal was associated with a comparatively high chance of winning office, in 2014 that chance is, statistically speaking, zero. This trajectory is the opposite of what might be expected given the citizenship politics in Latvia during this period. As the greatest number of naturalizations happened in the mid-2000’s, by 2010 the number of ethnic Russians eligible to vote in Parliamentary elections was drastically higher than it was in 2006. This would suggest that the natural constituency for ethnic Russian appeals was growing, even as the likelihood of candidates winning office by campaigning on a platform of Russian ethnic appeals was shrinking. It also suggests that even though some voters are still supporting parties which campaign as ethnic Russian champions—as seen in Figure 7.1—those parties which do so are relegated to the fringes of Latvian politics; those appeals are not able to actually elect anyone to office. This is in stark contrast to the general upward trend of electoral victory for Latvian ethnic parties and parties which do not explicitly identify as ethnic. The theory predicts that over time, candidates should have incentives to adjust their strategy depending on how they started. Electoral experience should demonstrate that non-ethnic and ethnic majority appeals are easier paths to electoral victory than ethnic minority Russian appeals. These patterns are evident in the data. To test these predictions outlined in 7.3, 7.4, and 7.5, I look specifically at those candidates who ran in multiple elections. Pooling all this data allows for an estimation of the over-time trajectories of candidate strategies and outcomes. The data suggest that revising one’s position on ethnic appeals is quite common among Lat- vian parliamentary candidates. Of the 1,047 individuals who appeared on party lists in more than one consecutive election in the period under study, 435 of them have changed their position regarding ethnic identification at least once. These changes happen either when a candidate

217 Candidates Changing Ethnic Appeals 0.5 1 −

C 0.4

0.3 Appeal at

≠ 0.2 C 0.1 Probability that Appeal

at0.0 Time Russian Latvian Non-ethnic Appeal at Time C

Figure 7.3: Likelihood of Candidate Changing Ethnic Appeal in Latvia 1998-2014

switches parties—moving from a non-ethnic party to an ethnic party, for example—or when a candidate stays within a party which changes its position on ethnic representation. To identify patterns in these switches, I regress whether or not a candidate changed her ethnic identiﬁcation

between elections C and C 1 on the indicators of the type of appeal made at election C. The pre- − dicted probabilities, plotted in Figure 7.3, are estimates of the average likelihood of a candidate making a specific type of appeal during an election after having made a different type of approval in the contest before. In other words, the predicted probabilities indicate the likelihood that a candidates will abandon a previous position in order to adopt either an ethnic Russian position, an ethnic Latvian position, or a non-ethnic position. Overall, an estimated 11.21% of candidates making appeals to Russians changed their party platforms in order to appeal directly to ethnic Russians. This is a much lower figure than the estimated 27% of candidates making Latvian appeals, who had switched their position in order to represent ethnic Latvians. The highest rate of change is in those making non-ethnic appeals. Pooling over all the electoral contests under study, the model estimates that 42.95% of the candidates changed their platform to non-ethnic from something else. These findings may be more intuitively understood by looking at the absolute number of candidates changing their political messaging between elections. I show these in Table 7.4.

The columns show the type of appeal made at election C, while the rows show the appeal made

218 at C 1 for all candidates who contested more than one consecutive election. The absolute − numbers should not be compared directly, as the total number of candidates in each group is not even distributed. There are fewer ethnic Russians than ethnic Latvians in the country, and as a result the overwhelming majority of candidates in Latvia are those which make ethnic Latvian or non-ethnic proposals to the electorate. Nevertheless, the bulk of the candidates are concentrated along the diagonal. This suggests that the majority of candidates stay consistent in their ethnic messaging between elections. Looking at candidates who advocated an explicitly ethnic Russian platform, 20 out of a total of 223 candidates, or 9.0% had previously advocated a multi-ethnic platform. However, among ethnic Latvian candidates, this number is 196 out of 736, or 26.6%. This comports with the general findings of the predicted probabilities derived from the logit model—in fact, both these absolute numbers are within the confidence intervals of the predicted probabilities. It suggests that candidates change to ethnic Latvian identification at a much greater rate than they do to ethnic Russian identification. Perhaps the most surprising finding is the non- zero number of candidates who switched between the two types of ethnic communities. These are the 10 candidates that started campaigning on platforms representing ethnic Russians and then changed to represent ethnic Latvians, and the 5 candidates who started representing ethnic Latvians only to switch to Russians—a complete about-face of ethnic campaigning. There are a few possible explanations for this apparent contradiction. The first is that these shifts may be driven by tokenism. Ethnic candidates of one group may be convinced to join the ballot of parties advocating for the other group’s communal interests to lend credibility to claims of moderation and fairness.3 The ethnic switching could also be the result of candidate desperation or strategic miscalculation. The findings above are constrained to the 1,047 candidates who contested more than one

3In interviews with Latvian political observers, voters, and elected officials, multiple people told me that tokenism was definitely present in Latvian politics. I was told multiple times that ethnic outsider candidates could possibly arrange for lucrative careers in the party if they were willing to put themselves on the ballot to indicate the party’s “enlightened” position, both to voters, and to European Union observers concerned about ethnic discrimination. It is also possible that some ethnic Russian candidates and voters may actually desire assimilation into an ethnonationalist Latvian state. One ethnically Russian Latvian candidate from an otherwise mostly ethnic Latvian political party told me during an interview that they were fine with members of their party arguing that ethnic Russians needed no linguistic, cultural, or employment protections from the state. In their view, Latvia was a state for the Latvian nation, and ethnic Russians should either assimilate, accept their status as “guests,” or leave. Such a view is hardly typical, and never once did I encounter an ethnic Latvian candidate willing to make similar arguments about the state protecting ethnic Russian interests, but these rare sentiments may help account for this negligible, but nevertheless present, rate of switching between ethnic parties of two different groups.

219 Appeal made at C Russian Latvian Non-ethnic

1 Russian 198 10 37 − C

Latvian 5 530 265 Appeal at

Non-ethnic 20 196 401

Figure 7.4: Count of Candidates who Changed Ethnic Identification between Elections in Latvia 1998-2014 election in a row. While these repeat contenders are the career politicians who make up the Latvian political class, most candidates do not fall into this category. There are 3,760 candidates who contested only a single election in the period under study. In a party list system like Latvia’s, parties may believe it is in their best interest to fill out the entire list of open slots, to maximize their ability to take as many seats as possible. No party has ever won an outright majority in Latvia, and in an electoral system with 100 seats and as many as 20 parties contesting elections, it is extremely unlikely for a single party to need a full 100 candidates on the ballot. Many parties put 100 candidates on their ballot so that it is mathematically possible, albeit highly unlikely for the party to sweep all seats in Parliament. Some parties even put more than 100 candidates on the ballot, suggesting that candidates may be seeking to give voters choices on the open-list ballot. While the majority of Latvia’s parliamentary seats are won by repeat candidates, one- time candidates represent a significant portion of Latvian elected officials. A full 30.33% of the parliamentary seats were won by candidates who had never contested elections before. While many elected officials are first-time candidates, most first-time candidates do not win election: only 5.1% of these candidates actually won a seat. Most of these candidates are placed extremely low on the party list, and likely enter with the full knowledge that will never be seated in the parliament.4

4In interviews, one mid-level party oﬃcial expressed embarrassment at the quality of the lower tier of their party’s candidates, and admitted that sometimes it was hard to ﬁnd candidates who were willing to sign their name

220 First Time Candidates by Appeal Type

0.42 0.09 V1 Russian Latvian Non-ethnic 0.49

Figure 7.5: New Candidate Entry in Latvia by Appeal Type 1998-2014

The data indicate that new candidates contesting elections follow the same trends with regards to ethnic campaigning as those candidates which switch their affiliation, only amplified. The pie chart in figure 7.5 indicates the proportion of new candidates who joined parties making each type of appeal. Candidates making appeals to ethnic Russians constitute 9.22% of first-time candidates, while candidates making appeals to ethnic Latvians constitute 42.05% of the new candidates. This is hardly proportionate to the ratio of Latvians to Russians in the population. There are 2.41 ethnic Latvians for every Russian in Latvia, but a new candidate is 4.55 times more likely to espouse an ethnic Latvian agenda than an ethnic Russian agenda. The ratio is similar for first time candidates espousing non-ethnic party platforms, who comprised 48.72% of all new candidates in the data set.5 Altogether the data show divergent trends between the various types of appeals. Support for ethnic Russian parties diminishes until it is statistically zero, while the proportion of winning candidates making ethnic Latvian and non-ethnic parties increases. Candidates joining parties and contesting elections for the first time are more likely to join non-ethnic parties or Latvian to the party platform, but wouldn’t embarrass the party with their lack of qualifications. 5Descriptive statistics are presented here because a logit analysis is not possible. The dependent variable in logit model to measure new candidate entry would be whether or not a candidate decided to contest an election in a given year. However, this is impossible to observe, since that would require us to identify all the “potential” candidates who chose not to seek election. Since we observe only those candidates who did enter the party lists, it is only possible to describe the relative frequency with which they attach their names to the three types of platforms described here.

221 parties than they are Russian parties. Candidates contesting elections repeatedly are more likely to abandon Russian platforms for non-ethnic platforms, and abandon non-ethnic platforms for Latvian ones. Over time, the party system converges on an ethnic cleavage that should really be understood as “ethnic Latvian versus non-ethnic,” rather than “Latvian versus Russian.”

7.4 Ethnic Position Change in Practice

The story of how ethnic appeals to Russians became relatively marginal in Latvia is largely the story of Harmony. Harmony is a political party which today has the support of most of the country’s ethnic Russian community. Ostensibly a multi-ethnic social democratic party, it has leveraged its near-monopoly of Russian-speaking voter support to become the largest single party in Latvia. It has done this partially by absorbing many of the smaller, more disparate parties which campaigned as ethnic Russian parties into a single organization. Harmony in its current form was founded in 2005, as an electoral alliance of factions consisting of the Peoples’ Harmony Party, the Latvian Socialist Party, the New Centre, the Social Democratic Party and the Daugavpils City Party (European Election Database 2018). Not all of these had been national- level parties, nor did they all campaign on ethnic issues. The Daugavpils City Party, especially, had only contested local elections in the city of Daugavpils, the country’s second-largest and the regional capital of Latgale, where a large population of Russian-speakers live. The Centre Party and the Social Democrats had large bases of support in Russian-speaking regions of Latvia, but campaigned on non-ethnic economic platforms. Much of their support from ethnic Russians may have come from nostalgia for Soviet-era stability, but the parties ostensibly were multi- ethnic and did not target only Russian supporters. The Latvian Socialist Party was a communist successor to the Latvian Communist Party which had been the only legal party under Soviet rule. That party did campaign explicitly to Russians, mentioning language rights and minority issues in previous platforms in terms which echoed the positions and propaganda of the Interfront in the 1980’s and 1990’s. The Peoples’ Harmony Party had also been previously connected with a political coalition that positioned itself as a group of minority rights advocates campaigning on legal and political protections for Russians in Latvia.

222 At its first election in 2006, the new party campaigned on a very moderate platform of Rus- sian rights. In its platform, the party explicitly stated its positions to protect non-citizen Slavic compatriots, and to foster better relations with the Russian Federation. The party’s appeal was directed mostly to Russians and Russian speakers, as is evident by the disproportionately high number of ethnic Russians appearing on the candidate list. Nevertheless, Harmony’s platform was more moderate than other parties appealing to ethnic Russian voters. In the same election, For Human Rights in a United Latvia campaigned not only on a platform of universal citizenship rights for all Latvian inhabitants, but also proposed that Russian be granted official language status, and that visa regimes should be liberalized to allow visa-free travel for Russians to visit family members in Estonia, Russia, and Belarus. In 2010, the divide between the two main Russian parties grew, as Harmony adopted an even more moderate platform while ForHRUL stayed consistent in its demands for communal rights for Russians. ForHRUL continued to espouse official status for Russian as its defining policy position, and argued that government bureaucrats and judges should be compelled to provide government and legal services to ethnic Russians in their native tongue. Harmony, meanwhile, eliminated all reference to non-citizens in its 2010 platform, and even included some nods to positions popular in the Latvian ethnic community. It proposed policies that encouraged the Latvian diaspora to return to Latvia proper (a position frequently espoused by right-wing politicians in Latvia as a way to ensure demographic dominance of the titular group but could nevertheless also apply to ethnic Russians who had left the country). It also stated that it was the responsibility of all Latvian inhabitants regardless of ethnicity to become proficient in the Latvian language, even while it stated the importance of recognizing the value of all who were “born or settled” in Latvia as an important part of a shared Latvian cultural history. Harmony adopted an even more moderate position for the early elections called in 2011. The party’s 2011 platform made only one mention of ethnic issues at all, arguing instead that economic crises were the most serious problems facing Latvia and that the party’s competitors would rather distract voters from those issues: “... political forces who represent both ethnic Latvian interests and the interests of minority peoples continue to emphasize national and historical issues and avoid social and economic problems.” ForHRUL, on the other hand, embraced

223 their ethnic identity, declaring, “as the party of the Latvian Russian Community, ForHRUL will guarantee equality of peoples and social justice.” By 2014, Harmony had abandoned entirely all mentions of ethnicity, nationality, or history from its platform. It no longer specifically claimed to be a multiethnic party or a champion of inter-ethnic competition, focused entirely on economic issues. It described itself exclusively as a social democratic party, and dedicated all of its 4,000 characters to economic growth, efficiency, and redistribution. ForHRUL, however, effectively disbanded and regrouped as a more extreme faction under the leadership of Tatjana Ždanoka calling itself the Russian Union of Latvia (RUL). The RUL platform in 2014 included not only an explicit declaration of protecting Russian interests and advocacy of official status for the Russian language in Latvia, but also a specific denunciation of fascism and Nazism, reminding voters of Soviet veterans’ contribution to the defeat of Hitler in World War II. During this period when the two parties diverged in terms of ideology, the clear winner at the polls was Harmony. At its first election in 2005, Harmony took 17 seats in parliament, to ForHRUL’s 6. This was a serious loss for ForHRUL, which had been the second largest party in Parliament going into the election, having won 25 seats in 2002. In 2010, the gap widened even more, with ForHRUL winning no seats to Harmony’s 29. The gap widened even further in 2011, when Harmony won 31 seats, yet ForHRUL’s again was unable to enter parliament. In 2014, Harmony remained the largest single party in the Latvian parliament, despite its loss of seven seats for a total of 24, but RUL failed to clear the threshold to win a seat.6 Despite the fact that both parties are often considered “Russian” by political observers both within Latvian and abroad, the more moderate stance of Harmony has undeniably had greater electoral success than ForHRUL’s more aggressive stance. While Harmony’s increased moderation and abandonment of ethnic claims was met with electoral success, ethnic Latvian parties’ route to victory often follow the opposite trajectory. One example is the Democratic Party “Saimnieks” (“DPS”), named after a merger between two other political organizations. The party took eighteen seats in the elections of 1995. In the tumultuous and fragmented environment of the early post-Soviet period, this made the party the

6While beyond the scope of this dissertation, the most recent Latvian elections in 2018 saw this trend continue. Harmony lost a single seat in 2018 to win only 23 seats, but again remained the largest single party in parliament, while the RUL again failed to meet the electoral threshold.

224 largest in parliament, and they formed a coalition government with six other parties. Given the

complexity of government formation with so many parties, the DPS supported Andris Šk, ele¯ for Prime Minister, a political independent running Latvia’s office of privatization, as an independent alternative to their own party leader. In that election, the party ran on an entirely non-ethnic platform. It presented to voters a classically liberal agenda, with respect for private property, advocacy of free markets, and support for entrepreneurship as a vehicle for economic development and job creation. The DPS platform made only two references to issues that could be remotely interpreted to reflect communal ethnic interest or national identities. The first, claiming that “whether they go east or west, there are no ‘good’ or ‘bad’ markets for Latvian exports.” The line seems to suggest that the party puts business above identity politics, arguing that Latvia could look to Europe and Russia equally as long as the decision was guided by economic interests. The party also mentions an important role for the Latvian language, stating, “[t]he opportunity to acquire knowledge must be guaranteed to everyone. The Latvian language will unite the population of the country into a family and a society.” While this does not imply the primacy of the Latvian language over Russian, it is tempered with language that suggests that the language policy is not necessarily intended to advantage ethnic Latvians, but instead to facilitate a more cohesive society. Despite its strong showing in 1995, the DPS did not do well in the elections of 1998, falling from the largest party in parliament to below the electoral threshold in a single electoral cycle. The big winner in 1998 was the People’s Party (“TP”) party created by Šk, ele¯ himself while Prime Minister. Šk, ele’s¯ new party leaned heavily on Latvian ethnonationalism in order to gain a plurality of seats in the Saiema. The party claimed that, “[t]he future of Latvia rests on healthy and happy children who speak the Latvian language properly, and are not ashamed to call themselves citizens of this country.” It goes on to state, that the family is the most important institution in Latvian society because it is the family that “guarantees the existence of Latvian traditions and Latvian culture.” The TP was able to build on its success in 1998, and remained a strong contender in the 2002 elections, as well, holding on to 20 of the 24 seats in had won previously. But during that time, many former DPS officials and candidates left their party to join the TP. This included Andris Ameriks, who had been the head of the Democratic Party half of

225 the DPS and remained a leader after the merger with Saimnieks. The TP would continue to be a major player in Latvian elections and eventually merged with Latvian Way. At every election it maintained its support for ethnic Latvian interests and campaigning as an ethnic Latvian party. This example shows how moderate leaders can take on more extreme positions through the interactive process of elections and party reformation. The leaders of the largest electoral party in 1995 had by 2002 joined a newer party which was not only more successful over a longer period of time, but also took a more extreme position on the representation of ethnic Latvian interests. Other DPS party leaders took similar paths after DPS ceased to be electorally viable albeit by joining other parties. Ilmars¯ Bišers, for example, who had worked closely with Ameriks in the DPS, joined the Centrs Party following the DPS collapse. The name “Centrs” may convey a moderate, centrist political ideology, but the party nevertheless used some relatively extreme language with regards to ethnic issues in the 2002 campaign. Its platform that year portrayed the party as a force protecting the Latvian people, and one which would help protect ethnic Latvian culture and traditions. It assured voters that if elected it would radically alter negotiating strategy with the EU, convincing the leadership in Brussels that Latvia’s unique historical context (i.e., occupation by the Soviet Union) was a special circumstance that did not subject Latvia to EU laws on human rights and political participation of minorities. It also argued that it would spend money to preserve traditional Latvian culture, and pass laws that prevented non-citizens and foreigners from owning property. These typical examples show that even when parties like the DPS fold after electoral failure, party members and leadership can ﬁnd ways to stay in Latvian politics by joining parties with more extreme positions than they held before. As another example, consider the Latvian Unity Party. In 1995, the party won eight seats in the parliament. This was not an inconsequential victory: in the extremely fragmented environment of the early post-communist elections, the largest party in parliament only had 18 seats total. The party that year presented a left-of-center platform that promised to end Soviet-style economic dirigisme, crackdown on corruption, increase education and healthcare spending, and protect human rights. The platform struck a conciliatory note with the Russian minority, ensuring that all minorities living within Latvia would enjoy “internationally recognized human

226 rights” and “the preservation of their own national identity.” The party advocated a strong, independent, and sovereign Latvian state, combined with a neutral foreign policy, which they argued would best allow Latvia to take advantage of its geopolitical strengths as a port between east and west. The party leader was Alberts Kauls, a prominent former Soviet official who had directed a series of state-owned agricultural concerns in Latvia during the Soviet period. Many other candidates in the party at this time were also former Soviet officials, leading Bugajski (2002) to classify it as a reformed communist party. The party’s relatively conciliatory tone towards ethnic interests may reflect this Soviet legacy and communist sympathy.

Although the party was successful enough to be represented in the Šk, ele¯ government, cohesion broke down while in government, and many deputies left the party before the 1998 elections. The drain caused the party to win no seats in parliament that year. Many of the party’s candidates and members, however, left to join the Latvian Farmers’ Union (LZS) (Smith- Sivertsen 2004). The common link shared between these two parties was the LZS’s focus on agrarian politics: many of the Soviet officials in the LVP had been in linked to Latvian agricultural production during the Soviet era, and had a natural constituency in the rural areas far from Riga. In 1998, the LZS platform echoed much of the conciliatory tone its members had voiced in the LVP. It mainly reflected the interests of those that lived outside the capital: a strong emphasis was placed on rural development and farming, as well as fishing, shipbuilding, and forestry. The party advocated for the gradual integration of non-ethnic Latvians into society “on the basis of volunteerism and loyalty.” The platform stated not that Latvian should be the official language, but also said that funds should be set aside for free Latvian language education for all those who did not speak the language. This moderate position changed in 2002, though, when the LZS—now campaigning as the Union of Greens and Farmers (ZZS)—moved to the right. The party argued that “only the Latvian people have the right to determine the future of Latvia,” and stated their goal to create a country with “the Latvian language as the only language, and the Latvian culture as dominant.” This advocacy of ethnic Latvian interests continued in 2006, when the party argued that it would reform the education system to foster a “patriotic” education. Positions like these were present in every ZZS platform onward, such as 2010, when the party said that traditional Latvian values

227 where the only “ethical and aesthetic basis" for the country; in 2011, when it argued that Latvian traditions were the most solid work ethic, and in 2014, when it stated that rural and traditional Latvian culture should be preserved with state funding and legal protection. This is not to argue that the ZZS is an extreme party, or even an overtly nationalist one. Given the party’s pre-communist origins, it could be considered comparable to other twentieth century agrarian parties of Northern Europe, such as Denmark’s Farmers’ Party, or Sweden’s Farmers’ League. Auers (2012) considers the party a curious deviation from the more commonly observed pattern of the modern Green movement in Western Europe—an ecological party ﬁnanced by an oil oligarch appealing to traditional values and Euroscepticism. In Latvia, the ZZS is often described as a centrist party. This may explain why half of the post-Soviet Pres- idents of Latvia have come from the ZZS. But in its relationship to ethnic appeals and communal interests, the ZZS shows a trajectory of incorporating more appeals to the ethnic majority group over time. Despite the fact that much of its leadership was associated with the Soviet Russian- dominated regime, and the fact that the party initially advocated multiculturalism for the Latvian state, the party ultimately adopted a harsher tone against ethnic and linguistic minorities, and has maintained that stance ever since.

7.5 Conclusion

Following the pattern of other recently democratized countries, Latvia has seen its political party system evolve as democratic institutions consolidate. Both elites and voters demonstrate forms of political learning, and parties have adapted the policies they propose to voters. But the way parties use ethnic cues to communicate has not been the same between ethnic groups. Ethnic appeals to the majority group have become a more common way to win elections over time, with a greater percentage of winning candidates making appeals today than they did in the 1990’s. Ethnic appeals to the minority group, however, have essentially vanished from the Latvian political system. While ethnic appeals to Russians and Russian-speaking Slavs were common in the immediate aftermath of the Soviet collapse, such appeals today are relegated to the fringes, essentially abandoned by mainstream politicians.

228 These results should not be taken as a sign that only ethnic Latvians care about ethnic issues. Quite the contrary: ethnicity remains one of the most salient social divisions in Latvia today, and ethnic distinctions have a significant impact on day-to-day life in Latvia. Nor is it fair to say that ethnic divisions are not relevant for party attachments. Harmony obtains the support of most of Latvia’s ethnic Russians, leading many international observers to refer to it as a “Russian Party.” But when it comes to making political decisions about how to employ ethnicity to mobilize voters, the two sides have different incentives to appeal directly to coethnics. For ethnic Russians, the costs of making ethnic appeals to voters simply outweigh the benefits. Over the past several decades, politicians have learned that appealing directly to Russians may help mobilize Russian support, but exacts a heavy cost in excluding the majority group. Appealing to ethnic majority Latvians, however, has a much less serious impact on politicians’ electoral prospects. In this sense, it is fair to say that the ethnic divide in Latvia is asymmetric: ethnic majority Latvians are much freer to use ethnic appeals when campaigning and building ties to voters than Russians are. As a result, it is not really fair to say that the primary divide in Latvia is not really between ethnic Latvians and ethnic Russians, but rather between majority Latvian ethnic nationalists and non-ethnic civic nationalists. These results should also be put into context, given the scope conditions imposed by Latvia’s institutional environment. Latvia has no guaranteed representation for ethnic minority groups, nor local autonomy, nor majority-minority electoral laws. The only control ethnic minorities have over policies is their ability to vote in free and fair elections. In this sense, Latvia is the polar opposite of consociational countries like Bosnia. The results suggest that absent explicit institutional protections, ethnic minority groups are likely to be represented by non-ethnic, civic nationalist groups which campaign on issues that may ostensibly appeal to voters of all groups. The results from the Bosnian analyses in Chapter 5 suggest the possibility of an unobserved counterfactual: if Latvia had adopted a consociational structure which granted minority’s some degree of self-rule and guaranteed representation rather than restoring the inter-war constitution in the early 1990’s, it is quite likely that ethnic Russian voters would today be represented by explicitly ethnic parties, rather than the declared non-ethnic social democratic parties which they support today.

229 8

Conclusion

In a democracy, elections are the mechanisms which translate citizen preferences to political leadership. To understand how democracy functions in an ethnically divided society, it is important to know how voters and candidates behave during elections in diverse environments. The politicization of ethnicity—wherein candidates get elected by appealing to coethnics to represent communal ethnic interests in government—is ultimately the result of choices made by political actors in pursuit of concrete goals. Whether or not ethnic mobilization is a viable and rewarding strategy is the result of institutional and demographic variables which may not affect all groups equally. This argument rejects the primordialist notion that ethnicity is a “given” of human existence.1 I do not argue that ethnicity is unimportant or inconsequential—in fact, it is vitally important to much of political life. But to understand the relationship between ethnicity and democracy, we need to appreciate the ways in which ethnic identities can help political actors achieve practical ends. “Ethnic politics” is just as political as it is ethnic, and even those identities most fundamental to an individual’s sense of self may very well be ignored if it is not helpful in achieving desired goals. We absolutely should examine the way that identities are formed and the significance that people attach to them, but we should also closely examine how those identities become politically relevant through the decisions of self-interested individuals. A strong personal identity is not a sufficient condition for political organization. If that were the case, the parliaments of the world would most likely be dominated by parties representing

1For details, see Geertz (1963), Hale (2004), Chandra (2012a), and Goode & Stroup (2015).

230 mothers and fathers, not workers and owners. The key to understanding when an ethnic identity becomes politicized is in appreciating the way it divides people into insiders and outsiders, and when such divisions are useful. Because ethnic labels are slow to change and defined by characteristics beyond an individual’s ability to alter, they can produce social dividing lines that are not easily manipulated for short-term gain. Even if those divisions are completely arbitrary—and in many contexts ethnic identities do seem to be based on relatively minor distinctions—they can still be instrumentally useful in organizing political life. When Bosnian voters vote for ethnic parties and non-ethnic parties in the same election, or when Latvian voters become more or less ethnic in the span of just a few years, we should not infer that those voters have radically changed the way they perceive themselves and their place in the world. Rather, what is changing is their relationship to their ethnic identities and their political goals. To understand that process, I have sought to disentangle identity and action. I start from the position that people do not have control over where they were born, who their parents are, what part of the world they come from, or what language is spoken in their community. But they do have control over how they integrate those background circumstances into their political lives. When they run for office, they can choose what they say to convince voters to support them. They can choose who they court for votes, and who they antagonize. And when they vote, they can choose who to support. These decisions may not change the nature of their ethnic identities in the short-term, but they do ultimately determine which identities become political relevant and which do not. I have shown in this dissertation that substantial variation exists in these decisions, and that such variation is systematic. When ethnic identities separate people into groups that are able to win office and likely to deliver benefits to their supporters, political actors take observable steps to politicize ethnicity. Candidates emphasize their own ethnic identities over individual characteristics, and voters elect them to office. When ethnic identities are not likely to result in useful political organizations, then candidates win by campaigning on other, non-ethnic identities. The outcomes I have documented in this dissertation are the result of individual choices, but they are heavily influenced by background situation. Institutional and demographic context structure the relationship between ethnic groups, and create the incentives facing political actors who make these decisions.

231 This approach contributes to the literature on ethnic politics and democracy in diverse societies, partly by drawing attention to our understanding of the subject of “ethnic politics” itself. Because ethnic mobilization is not equally effective for all groups, the democratic processes which translate societal diversity into political representation do not function the same way for everybody. To understand politics in ethnically divided societies, we need to understand that outcomes are the result of the relationship between groups, and ethnic representation may be qualitatively different in different environments. One important conclusion is that approaches which examine ethnic minority political issues as independent from majority group politics may ultimately ignore the underlying factors which provide opportunities for competition and cooperation between the groups. The two cases in this dissertation show how outcomes of ethnic tension may be produced by very different underlying causes. Bosnia and Latvia both have substantial social tension related to minority group representation. In Bosnia, majority-group ethnic Bosniaks and the much smaller ethnic Croat population group vote for separate parties, with very little cross-ethnic voting taking place. Likewise, majority group ethnic Latvians very rarely vote for the same parties as minority-group ethnic Russians. Once might conclude that both countries are similar and represent similar dynamics of ethnic mobilization. But the choices made by the actors that lead to that outcome are drastically different. Croats vote for ethnic Croat parties in Bosnia, choosing explicitly ethnic parties at a higher rate than ethnic Bosniaks do. Croats have a choice between ethnic and non-ethnic parties and largely choose to vote for the ethnic parties, producing an observable ethnic divide in the party system. Russians in Latvia, on the other hand, vote overwhelmingly for a party which claims a civic identity and rejects ethnic mobilization. The ethnic divide which separates Russians from ethnic Latvians is largely the result of ethnic mobilization in the majority group, which increasingly uses exclusionary rhetoric and policies of ethnic favoritism to exclude minority group voters. Both outcomes could accurately be described as “ethnic politics,” but the two are radically different. In Bosnia, the divide is produced by the choices of the minority group; in Latvia it is the majority group. This not just important for scholarly understandings of the relationship between ethnicity and politics, but also for practical issues of fairness, representation, and democratic accountability. One of the greatest concerns in ethnically divided democracies is that minority groups

232 face permanent exclusion from power. Ensuring minority group representation and inclusion is therefore a pressing policy issue directly related to the quality of life for large numbers of people around the world. But ethnic minority representation cannot be understood while ignoring the choices made by the majority group. This dissertation suggests that in most contexts, politicizing ethnicity ultimately benefits the majority group. Democracy is at its core a majoritarian system, where big groups win and small groups lose. Many institutions which empower minority groups and facilitate their election do so by converting them in some small way to majority groups, usually by changing the borders of the electorate or the level of governance. But doing so also changes the relationship between minority groups and majority groups by altering both group’s incentives to politicize ethnicity. This dissertation calls for a greater understanding of the incentives faced by both types of groups and how changes in one group’s calculations can affect the other group. In this concluding chapter, I reflect on the broader implications of this dissertation. I first discuss how the findings should be put into context of other studies of ethnic politics and democratizing countries, including the applicability of the results to other cases. I also discuss how a research design that relies on within-country variation and takes advantages of idiosyncratic cases may nevertheless further our understanding of ethnic politics in divided societies more broadly. I the closing section, I use the findings of this dissertation to discuss the outlook for ethnic voting and majority-group political mobilization in the world today, with particular attention to the increase in anti-minority rhetoric and majority group mobilization in the United States.

8.1 Implications for Future Studies

In this study, I have employed original data sets which coded political parties and candidates appeals to voters on ethnic grounds. These indicators reﬂect whether or not voters would interpret the party as representing voters from a speciﬁc ethnic group to the exclusion of others. It should be noted that in the two cases surveyed here, it is not unusual for candidates to employ explicit ethnic rhetoric in campaigning. Parties and politicians in both of these countries sometimes

233 view ethnic identities as so central to their political brand that the name of the ethnic group is used in the name of the party. But this is not the case everywhere. In some countries, appealing directly to a specific ethnic group is considered taboo, and is relatively rare. Exclusionary ethnic positions are often present in these places even if coded. The United States especially is notorious for “dog-whistle” politics, or candidates making proposals and statements intended to appeal to those who harbor racial or ethnically exclusionary political preferences while appearing innocuous to those who do not. By definition, these statements are intended to be interpreted in two different ways, intentionally obfuscating politicians’ true beliefs (Albertson 2015, Haney- López 2015). Applying this study’s methodological approach in these environments could be especially difficult, since whether a party is ethnic or not is intentionally disguised. Moreover, some scholars have also noted that in contexts where ethnic inequality is deeply entrenched, and societal institutions favor one group over another, simply supporting the status quo can effectively be taking a position on ethnic issues. Given certain contexts and reputational resources, some candidates are capable of making ethnic appeals using a language of equality that is much less straightforward and easily identified than those cases in this dissertation (Mendelberg 2001). Further complicating the prospects of exporting this approach to other regions or cases is that many countries have placed legal bans on ethnic parties (Bogaards, Basedau & Hartmann 2010). In these situations, parties which are founded explicitly to represent a specific ethnic group are illegal. It would be very difficult to observe an ethnic party in these contexts. Yet in many of these countries parties which represent the interests of ethnic minority groups do exist and often win elections by mobilizing coethnics. Many scholars treat these parties as no different from ethnic parties in other countries where the ban does not exist.2 The measurement approach used in this dissertation would not be appropriate. In order to avoid being banned, these parties must avoid ethnic rhetoric which would clearly establish their position as representing only a single ethnic group. All parties would therefore be deemed to be non-ethnic using the standards employed here. These situations illustrate the necessity of local contextual knowledge to the study of ethnic politics. The design employed here relies heavily on ethnographic explorations of the history

2See, e.g., Alonso & Ruiz-Ruﬁno (2007).

234 and the political language of the countries and the ethnic communities under study. This makes it possible to document how candidates make ethnic appeals and how voters are likely to interpret them. This local knowledge is necessary to identifying and classifying the types of strategies parties use to mobilize supporters. Local context is vitally important, and blindly applying the standards of one country to another will inevitably be inappropriate and misleading. In Bosnia, a fleur-de-lys is a polarizing ethnic symbol. It is not in France, but it can be in Canada. Building a statue to commemorate veterans of World War II is likewise a polarizing ethnic demonstration in Latvia. It is most often a message of multi-ethnic unity in the United States, but an ethnic provocation in Japan. The difference in each of these cases is how the groups perceive them and how historical factors have shaped the ethnic identities over the long-term. If two groups have the same relationship to a given statement, idea, or symbol, then its use in a political campaign does not signal anything about the candidate’s allegiance to one group or the other. Since the mechanism theorized in this dissertation works through individual motivations and understandings contingent on social environment, context matters. While dog whistles and ethnic party bans make the work of measuring ethnic appeals much more complicated, they do not make it impossible. They do, however, require greater attention to be paid to identifying how candidates employ ethnic identities in the process of winning elections. The guiding principle behind the data collection of this dissertation was sensitivity to local context and within-country variation. Nevertheless, the findings and the explanatory mechanisms can be seen to apply to other cases outside this dissertation. The clearest extension of this study is to other recently democratized countries of Central and Eastern Europe. The Latvian case discussed in Chapters 6 and 7, for example, can shed light on differences observed in the party systems today in the other post-Soviet Baltic states. In Estonia, for example, a very similar demographic and institutional context has produced a very similar result. Estonia is 68.7% ethnic Estonian, and 29.6% Russian-speaking, very close to Latvia’s 62.07% majority (Central Intelligence Agency 2017). Like their Latvian neighbors, Estonia is also a highly centralized state using a proportional representation electoral system. As in Latvia, Russian-speaking Es- tonians are largely represented by a non-ethnic party. Following the Soviet collapse, the Centre Party—founded by the half-Estonian, half-Russian Soviet official Edgar Savisaar—has come to

235 enjoy the overwhelming majority of Russian-speaking Estonians’ votes. This party, a direct successor to the Estonian Popular Front which led the movement for independence from the Soviet Union, defines itself as a centrist populist organization, and also has attracted a not insignificant number of ethnic Estonian supporters. Throughout the post-Soviet period, the party’s moderate stance has defeated more explicitly ethnic parties, including the Party of Russians in Estonia, in order to win the support of most ethnic Russians within the country. Following the 2016 election, the Centre Party entered into government and took the position of Prime Minister, largely due to the steps taken by party leadership to distance themselves from the party’s ethnic Rus- sian identity. The party removed Savisaar from his position as party leader, and “froze” their agreement with Putin’s United Russia (Martyn-Hemphill 2016a, Vanajuur 2019). Meanwhile, some leaders among the ethnic Estonian majority have enacted policies designed to entrench the rights of ethnic Estonians. Like Latvia, Estonia passed citizenship policies in the 1990’s which legally determined that many ethnic Russians were invaders, not citizens, and enacted laws throughout the 1990’s and 2000’s to legally control the Russian language and prohibit its public use (Hughes 2005). Estonian ethnic identity plays a prominent role in the ideology of several major Estonian political parties, including the Conservative People’s Party of Estonia and Pro Patria (Bennich-Björkman & Johansson 2012, Nakai 2014). There are important differences between Latvia and Estonia, and the similarity of the two cases should not be overemphasized. Nevertheless, in both cases there is a very similar pattern of ethnic mobilization taking place in the ethnic majority group, while minority group members rally behind ostensibly non-ethnic parties. Lithuania, on the other hand, shows a different outcome. The ethnic Lithuanian majority is much more solid, and has been since even during the Soviet period. At independence, ethnic Russians comprised only 9.2% of the population. Because of Lithuania’s history as part of the Polish-Lithuanian commonwealth, ethnic Poles presented an ethnic minority nearly as large as the Russian community, accounting for 7.0%. Both these groups founded their own political parties in the mid-1990’s in the early post-Soviet period: Lithuanian Polish Electoral Action, and the Lithuanian Russian Union (Novagrockiene˙ 2001). But the Lithuanian electoral system is quite different from the other Baltic states: Lithuania uses a mixed system where half of the

236 seats are elected through PR, and the other half elected by SMDP. Moreover, the differences in the size of the minority communities in Lithuania created drastically lower incentives for the ethnic majority group to pursue ethnically exclusive strategies. These key institutional and demographic differences produce very different outcomes in Lithuania. Among the majority group, ethnic rhetoric is much more sedate in Lithuania than its Baltic neighbors. Lithuania granted citizenship to all residents at the time of independence from the Soviet Union regardless of ethnicity or family origin, and punitive policies towards non-ethnic Lithuanians has been relegated to the fringes of Lithuanian political discourse. Because the ethnic Lithuanian majority was so much more secure in Lithuania than it was in other Baltic states, and the benefits of EU and NATO membership were always so high, nationalists largely abandoned the incendiary ethnic rhetoric seen elsewhere in favor of support for minority rights and western integration (Clark 2006). Ethnic minorities in Lithuania also act differently than minorities in other states. The Lithua- nian Russian Union, while existing as a formal organization, has hardly ever run as its own party. In almost every election since independence, it has formed a pre-electoral coalition with larger parties of non-Russians, usually either Socialists, or the ethnic Polish minority. When doing so, it usually downplays its ethnic nature, running on a joint platform stressing universal issues of human rights and regional development (when in coalition with the Poles), or economic fairness (when running with the Socialists) (Frejut˙ e-Rakauskien˙ e˙ 2011). Lithuanian Polish Electoral Ac- tion, on the other hand has maintained its independence, and explicitly campaigns directly on Polish identity more frequently. Its signature issues in many elections have been repealing the legally-mandated Lithuanian spelling of Polish names,3 and the return of territory taken from Poland at various points in history and given to Lithuania (Sabanovic N.d.). The difference between the Russian and Lithuanian political parties may be partially attributable to the interaction of demographics and institutions. Poles make up a majority of the population in several municipalities in Lithuania along the southeast border of the country. Their geographic concentration

3For example, the common Polish name “Wiśniewski” cannot be legally listed on a Lithuanian passport. The letter “W” does not exist in the Lithuanian alphabet, nor is “-ski” a grammatically correct adjectival ending in Lithuanian. Ethnically Polish Lithuanians must use the Lithuanian form “Vyšniauskas” as their legal name. Latvia has a similar policy (the legal spelling would be “Višn, evskis”) yet the naming issue among ethnic Russian parties in Latvia is very rarely taken up in party platforms.

237 gives them an advantage in the SMDP seat elections, such as in 2000 when they failed to clear the threshold for representation in the PR seats, but took two seats in the single-member constituencies. In this sense, the Poles in Lithuania may in many places be like the Croats in Bosnia: a minority in the country as a whole, but a majority within their electoral districts, and therefore have increased incentives to support ethnic parties. Russians, however, are more geographically dispersed, and must de-emphasize their ethnic rhetoric and seek non-ethnic coalition partners. The findings of this dissertation may also help understand the varieties of ethnic politics beyond the immediate region. In Latin America, for instance, scholars have noted that when indigenous populations are mobilized by activists into democratic politics for the first time, they often do not translate their social movements into ethnically exclusive political parties. This surprising tendency goes against the expectations of the literature on social cleavages and party systems, as well as the experience of many first-wave European democracies. Instead, indigenous politicians in places like Bolivia and Ecuador are more likely to adopt multi-ethnic populism even as they rely on indigenous voters for electoral support (Van Cott 2005, Madrid 2012). The findings of this dissertation suggest that this phenomenon may be a result of the way that electoral institutions and demographics encourage indigenous leaders to moderate their ethnic positions to win votes from across the ethnic divide. Latin American countries often utilize majoritarian electoral systems which disadvantage smaller parties while disproportionately rewarding parties which obtain a plurality of votes. Evo Morales, the first indigenous president of Bolivia, has made his own ethnic Aymara identity an important part of his political brand, but not more so than his non-ethnic leftist policy positions. His campaign and subsequent presidential administrations have prioritized a socialist, anti-neoliberal, and anti-colonial agenda in addition to appeals to the country’s indigenous groups. This “soft” ethnic populism makes sense in the Bolivian context. Bolivia uses a two-round electoral system to elect the president, so a candidate must win a majority (or near to one) in order to take office. Morales’s Aymara ethnic group comprises roughly 10% of the Bolivian population—hardly enough to win an election which ultimately requires a majority. Given the fragmentation of the country’s indigenous population, a far more viable strategy is to build non-ethnic coalitions based on class and economic populism. These positions are likely to appeal to indigenous peoples of other groups while also

238 attracting support from lower-class mestizo voters. It is important not to overgeneralize with regards to other cases outside the region. Since the research design in this dissertation relies heavily on local context, further research into the electoral strategies and the decisions of the electorate is obviously required. Nevertheless, the findings do suggest a framework to approach the variation in elite incentives imposed by democratic institutions around the world. They suggest a mechanism that could produce both ethnically divided party systems like Trinidad and Tobago and Fiji, where leaders attract ethnic bases of support but nevertheless moderate their positions on ethnic representation (Wilson 2012, Fraenkel & Grofman 2006); and Turkey and Sri Lanka, where ethnic parties take antagonistic and even violent positions against ethnic outsiders (DeVotta 2004, Kirisci & Winrow 2013). Perhaps the most important implication of the findings in this dissertation is that research designs which rest on country-level variation may miss some important dynamics of ethnic conflict and the representation of ethnic groups. Much of the literature is built around dichotomous questions explained by country-level variables: Does PR encourage ethnic voting (Huber 2012)? Do ethnic party systems encourage ethnic outbidding (Mitchell, Evans & O’Leary 2009)? Does federalism encourage separatism (Brancati 2006)? Do ethnic quotas enhance representation (Dunning & Nilekani 2013)? These are important questions and these studies have produced important insights. This dissertation suggests that we may advance this line of study by appreciating that majority and minority groups react differently to different institutional changes. Eth- nic conflict and representation are ultimately about the relationship between groups: outcomes are shaped by the independent decisions of majority and minority group and the incentives to which they respond. Changing from a majoritarian system to a proportional system, or creating a special district or reserved parliamentary seats for a specific minority group may enhance a group’s capacity for representation, but it may also remove any incentive the majority group has to court support from ethnic outsiders. These asymmetrical effects may explain why attempts to mitigate ethnic conflict through institutional engineering are so often frustrated in practice: changes intended to alter outcomes for one group produce unintended changes in incentives for another. Even within the same country the incentives for political actors may vary across groups in systematic ways. This effect should also inform studies which examine when minority groups

239 are most likely to be included in the political process.4 In addition to divergent effects between groups, this study suggests that groups can be represented in qualitatively different ways, which may produce different substantive outcomes. This study suggests that increasing the political participation of excluded ethnic groups can take a variety of forms, with different implications for party systems and elections. Increased minority political participation could increase the power of groups which seek to de-politicize ethnic identities and foster broader, multi-ethnic political alliances in parliament. It could also deepen ethnic divisions by enabling new ethnic parties to mobilize voters along ethnic lines while reinforcing ethnic cleavages. These are qualitatively different phenomenon, and cannot be explained by research designed that conceptualize ethnic political activity in dichotomous ways.

8.2 The Fate of Diverse Democracies

This dissertation also speaks to concerns of ethnic outbidding, polarization, and political extremism in ethnically divided societies. While this study does not propose a deterministic account of when ethnic parties will succeed, or when they will adopt extreme positions, it does suggest that incentives for ethnic political organization are not equal across groups. One important implication suggested by the findings in this dissertation is the asymmetric ability of one group to moderate the positions of the other. This dissertation finds that moderate positions and non-ethnic mobilization among the majority group is most likely when ethnic voting is not likely to deliver any practical benefits to group members. This would be the case when the majority is secure in its position, and does not face any serious threats to its dominance by minority groups. Minority groups, on the other hand, may be in a vulnerable position, but still unable to win elections by adopting extreme ethnic positions. The minority group faces a much more punitive backlash for ethnic mobilization than the majority group. As a result, the majority can force moderation on the minority, but the minority can only force extremism on the majority. The practical implication of this ethnic asymmetry is that the instinct to guarantee minority political involvement by mobilizing ethnic minorities using ethnic policy proposals may

4See, for example, Cameron, Epstein & O’Halloran (1996), Pande (2003), Givens & Maxwell (2012), Sobolewska (2013), and Dancygier (2017).

240 be counter-productive. By way of example, consider the Black Lives Matter movement in the United States. A 2017 survey found that 56% of the American public—and 57% of whites— believed that police officers in the United States were too quick to resort to force when engaging with suspects. The same survey found that the Black Lives Matter movement, originally founded in response to police violence against African Americans, was viewed favorably by only 43% of the public. This figure is the result of polarized attitudes along racial lines: 83% of black respondents supported BLM, while only 35% of whites did (Easley 2017). The survey suggests that use of racial cues in advocating for changes in policing may be counterproductive, and that white Americans may be put off from policy changes they would otherwise support by a fram- ing linked to ethnic minority populations. While “All Lives Matter” is a logical superset of the claim “Black Lives Matter,” the statement has become a common way to criticize the goals of BLM on social media. The opponents of BLM rarely question the fact that police violence is far more common against people of color; instead, they criticize the movement’s goals as racist, advocating for special rights for black citizens at the expense of other ethnic groups. “All Lives Matter” is often little more than a disingenuous rhetorical device to dismiss the valid concerns of African-American activists.5 But the example of Black Lives Matters shows why exclusive minority ethnic mobilization may ultimately be a losing battle in a democracy. The devastating impact of police killings in the United States has overwhelmingly fallen on non- white communities. It is therefore unpleasant, to say the least, to respond with a political strategy which places the concerns of white voters first and foremost. But surveys show that white and non-white respondents in the United States often agree in their distaste for police actions which fall disproportionately on minority citizens, such as racial profiling, excessive force, and civil asset forfeiture (Eakins 2016). Finding the policy positions supported by majority group voters and presenting them as non-ethnic issues may therefore ultimately be a more viable short-term strategy for implementing policy change. This should not be construed as support for the “color- blind” politics which often erases the experiences and valid concerns of minority communities, and has often entrenched racial disparities in the United States (Bonilla-Silva 2006). Rather, it is a plea to understand the nuanced ways that ethnic appeals ultimately benefit majority groups.

5See Carney (2016) for discussion.

241 From the perspective of this dissertation, racial inequality in the United States is partially sustained through the successful ethnic appeals of candidates to the white ethnic majority. In this light, ethnic appeals to a minority are doomed to failure, and the better strategy is to counter ethnic appeals with non-ethnic appeals. In a diverse society, the ethnic majority groups has the luxury of excluding ethnic outsiders from their political coalitions and still winning office. Simply by virtue of their demographic status, minorities must build coalitions that cross ethnic lines, or limit their influence only to areas where they themselves are a local majority. That majorities may have an easier time winning by making ethnic appeals may also help explain recent trends towards ethnic voting in developed democracies. It is clear that majority group ethnic voting is in resurgence at the moment in many parts of the western world, and a backlash against globalization has seen an increase in ethno-nationalism in many countries. Brexit in the United Kingdom passed in 2016, with support heavily divided between majority and minority ethnic groups. The “leave” vote was heavily concentrated in England in ethnically homogeneous areas, while the “remain” vote concentrated in ethnically diverse cities, and ethnic minority enclaves Scotland and Northern Ireland. In Hungary, Viktor Orbán’s Fidesz has won an overwhelmingly large parliamentary majority on appeals to protect Hungarians from ethnic outsiders, especially Arab refugees and European bureaucrats. Lega gained a plurality of seats in the Italian parliamentary elections of 2018 on an anti-Muslim and anti-immigrant platform, while France’s far-right Rassemblement National and Germany’s Alternative für Deutschland have shown increasing support among voters. In the United States, Donald Trump won the presidency in 2016 on a platform that broke long-standing American taboos against overt ethnic campaigning. This dissertation offers a tentative explanation of why majority-group ethnic campaigning is becoming more successful as the western world diversifies. I have argued that a growing ethnic minority population increases the benefits of ethnic redistribution to the majority group, without really threatening their ability to win elections through ethnic mobilization. New immigrants, refugees, or growing ethnic minority populations can increase labor market competition, in which case discriminatory policies against minority communities may help increase wages for majority group members. Imposing taxes on minority group members while structuring

242 programs to provide benefits only to majority group members may also be a way to increase services to constituents for politicians seeking reelection. At the same time, as long as the increase in minority populations is small enough not to threaten the ethnic majority’s hold on political power, it is virtually impossible for ethnic minorities to elect their own leaders to prevent such exploitative policies from being enacted. In this way, it is not surprising that the increased diversity of Western Europe brought on by the expansion of the European Union and the recent refugee crisis has empowered majority group ethnic politicians. Examine, for instance, the institutional and demographic environment of the United States. It is a presidential system, where the legislative and executive branch are elected through separate elections. It is also a federal system, where states and even cities often have a high degree of independence. At virtually every level of election, it employs a first-past-the-post system with two dominant political parties. In the US electoral system, geographical boundaries are crucial. Each state receives a number of electoral votes to choose the president relative to its population, and the upper house of the legislature is decided on the basis of territory, not population. De- mographically, the US has a clear majority ethnic group, with whites comprising 60.4% of the population (US Census Bureau 2018). Birthrate disparities and increased immigration suggest that the white majority in the United States is diminishing, and will likely last only until the mid-2040’s. However, the geographic dispersal of white people in the United States is hardly uniform. Thirty-seven of the fifty US states are proportionately whiter than the population of the US as a whole, and forty-five of them are majority white. Nearly one-third of all non-white people in the United States live in just two states: Texas and California.6 Many observers concluded that the future of American politics would see less of an emphasis on ethnic identity politics as the United States diversified. Following the defeat of Mitt Romney to Barack Obama in the 2012 election, the US Republican party released its famous “Autopsy”—officially known as the RNC “Growth and Opportunities Project” report. It concluded that the party’s reliance on white voters to the exclusion of other ethnic communities was dooming the party to irrelevance and perpetual defeat. The authors argued that the party “...need[s] to campaign among Hispanic, black, Asian, and gay Americans and demonstrate we

6My own calculations taken from US Census Bureau (2018).

243 care about them, too. We must recruit more candidates who come from minority communities. (Barbour, Bradshaw, Fleischer & Fonalledas 2013, p. 6)” The project also concluded that it needed to appeal to ethnic Hispanic voters by using non-ethnic rhetoric and policies, noting that, “...among the steps Republicans take in the Hispanic community and beyond, we must embrace and champion comprehensive immigration reform. If we do not, our Party’s appeal will continue to shrink to its core constituencies only. We also believe that comprehensive immigration reform is consistent with Republican economic policies that promote job growth and opportunity for all. (Barbour et al. 2013, p. 6)” Many members of the political establishment were therefore shocked by the victory of Don- ald Trump, who announced his candidacy by calling Mexican immigrants rapists.7 For a campaign known for a historically unprecedented lack of ideological consistency, Trump stayed remarkably constant in his anti-immigration rhetoric, rarely deviating from a strategy of courting white voters at the expense of all other racial and ethnic groups. Even on his signature issue of restrictions to immigration—the one issue which has deﬁned the Trump political agenda more than any other—Trump has shown a willingness to compromise on the basis of ethnicity, famously declaring that he would welcome immigrants from white European countries like Norway, preferring them to those that come from “shithole countries” in Africa, Latin America, and the Caribbean (Dawsey 2018). Trump’s propensity to antagonize an enormous segment of voters to court an overwhelmingly white base has shocked and dismayed not only academic observers, but members of Trump’s own party. His willingness to emphasize ethnic distinctions during his campaign even when doing so would compromise the working of the federal government was especially trou- bling. Trump famously attacked the independence of the judiciary by questioning the credentials of an Indiana-born federal judge for being “Mexican” (Kendall 2016), and undermined the armed forces tradition of veneration for fallen service members by attacking the surviving

7While the oft-quoted line from that speech is, “They’re bringing drugs. They’re bringing crime. They’re rapists. And some, I assume, are good people,” that part of the speech actually began with, “When Mexico sends its people, they’re not sending their best—they’re not sending you. They’re not sending you. (Burns 2015)” That the quote is racially divisive should not distract from the fact that the speech is also based on establishing distinctions between in-groups and out-groups. Trump makes it clear that the group of voters he is courting contains “you,” not “Mexicans,” making a very clear declaration of who he plans to exclude while in oﬃce. This is the very deﬁnition of an ethnic appeal used in this dissertation.

244 family of a Pakistani-American soldier who died in a car bombing in Iraq (Turnham 2016). The independence of the judiciary is enshrined in the US constitution and is considered an important institution in the workings of the American legal system. Likewise, respect for fallen soldiers is an important element in the way the US armed forces recruits soldiers and maintains military discipline. That Trump was willing to disregard these vitally important elements of American political life in order to make ethnically divisive appeals to voters was seen by many in both of America’s two major parties as an indication of being unfit for office. Moreover, many in Trump’s own Republican Party questioned it as a viable campaign strategy. Some in the party believed that Trump’s reliance on incendiary ethnic rhetoric even when it challenged respected American traditions would not only exclude the possibility of any non-white electoral support, but also alienate white moderate voters in crucial swing states. With less than a month before the general election, more than three dozen elected Republicans had called on Trump to stand aside and hand the presidential nomination to someone else, believing that with Trump at the top of the ballot the party was doomed to failure (Blake 2016). Clearly, this was not the case. While Trump failed to win a majority (or even a plurality) of votes, he nevertheless won the presidency. More than one commentator concluded that the election had finally dispelled the comforting but erroneous notion that in the United States, racism was disqualifying for office (Harris-Perry 2016). This dissertation suggests a framework to understand why the consistent appeals to white voters and the brazen dismissal of non-white votes may be in a winning strategy in the US today. By emphasizing his willingness to violate norms on ethnocentric rhetoric, Trump signals to voters that he will favor white voters at the expense of non-white ethnic outsiders. In this way, Trump’s divisive speeches are the American translation of the ethnic appeals used in the new democracies of Central and Eastern Europe. Trump describes Spanish-speaking immigrants as invaders using illegitimate chain migration to claim legal rights, the same way that far-right Latvian ethnic parties attack ethnic Russians for benefiting from an illegitimate Soviet occupation. Trump’s increasing public support for monuments dedicated to white supremacists like Robert E. Lee despite the extreme unpopularity of such figures in non-white communities8 is not dissimilar from the way Croat parties in Bosnia

8See How Donald Trump learned to love Confederate monuments (2017).

245 adopt the names of nineteenth-century radical nationalists to signal their appeal to one group at the expense of another. Just as Latvian ethnic parties portray ethnically Russian citizens as a threat to the sovereignty of an ethnically Latvian nation state, Trump publicly tells non-white legislators to “go back” to countries of which they are not citizens (Rascoe 2019). The language may be different, but the message is the same: a clear division between rightful and deserving insiders and threatening undeserving outsiders defined in ethnic terms. As the United States becomes more diverse, but nevertheless retains its white majority, the attractiveness of such a position may be increasing to white voters. The success of Trump’s divisive politics may not be an aberration, but the result of long-term trends. Trump’s rhetoric is an attempt to transition the Republican Party from an ideological party reflecting free-market liberalism and traditional conservative social values to an ethnic party benefiting whites. In an environment where whites makeup roughly 60% of the country, the benefits of a white ethnic party are extremely high to white voters. The prospect that non-whites could be excluded from receiving social services will not be provided to non-whites could be extremely lucrative to white voters. Such a policy would make 40% of the country pay taxes to subsidize the white majority while receiving none of the benefits. The Trump administration has already taken limited steps in this direction. Low-income predominately white areas in Appalachian coal country and the agricultural Midwest are the “forgotten” people that the administration promises to help with federal subsidies. Low-income predominately non-white areas like the late Elijah Cummings’s district in Baltimore are “rodent-infested” and “failed,” where federal funds would be wasted (Rascoe 2019). In this way, the federal government can effectively act as an inter-ethnic redistribution mechanism, taxing both black and white, but designing subsidies and budgets the disproportionately benefit an ethnic political base. While the benefits from ethnic redistribution are increasing as the US diversifies, American institutions ensure that the likelihood of white electoral victory changes very little. Non-white voters are disproportionately concentrated within specific states, and as such their ability to prevent a white ethnic party from getting elected is limited. Only in Nevada, Texas, New Mexico, California and Hawaii are non-Hispanic white voters a minority.9 Consider a hypothetical ethnic

9This is also true of the District of Columbia, Puerto Rico, American Samoa, the Northern Mariana Islands, the US Virgin Islands, and Guam, but these territories have no voting congressional representation.

246 census election, where ethnicity perfectly corresponds to voting behavior. In this election, a non-white ethnic party which could mobilize all non-white voters facing off against a white ethnic party mobilizing all white voters would secure only 10 senate seats, to the 90 won by the white party. In the Electoral College for the presidency, the race is closer, with the white ethnic party winning 430 Electoral College votes to the 108 votes won by the non-white rival. In this stylized example, white voters have extremely high benefits to electing ethnic representatives— in fact, the population breakdown of the US between whites and non-white is very close to the breakdown between Latvians and non-Latvians in Latvia. But the geographic nature of the US electoral system and the dispersal of the population eliminates any serious barriers to winning office for white ethnic parties. This lopsided balance is likely to get worse before it gets better in the United States. As the white majority diminishes, the benefits of ethnic representation for white voters will only increase. But unless the non-white population relocates into mostly white states, political parties and candidates will mostly face no new consequences for campaigning on platforms of coethnic favoritism towards whites. In this way, an increasingly diverse United States could actually see an increase in overtly racial campaigning. For those who are concerned by the increase in divisive ethnic and racial rhetoric, the findings here suggest long-term and short-term courses of action. In the long term, the US must make it a priority to end the gerrymandering and the institutional imbalance created by the territorial allocation of political power which in no way corresponds to the distribution of ethnic groups. Eliminating the Electoral College, for instance, would change the results of a pure ethnic census voting from an 80% to 20% contest in favor of ethnic whites to a 60% 40% contest. This would drastically reduce the incentives to campaign to white coethnics for the majority group, while in no way encouraging minority groups to campaign on ethnic grounds. Somewhat counterintuitively, the US should also rethink the extremely common practice of minority-majority districting. This ethnic gerrymandering is often intended to make it possible for voters from non-white groups to elect coethnic representatives despite being a minority of their state or local area. But the findings in this dissertation suggests that minority-majority districting encourages ethnic voting in both majority and minority groups, widening the gap between the two. Group- ing minority voters into their own district does eliminate their barriers to entry, making ethnic

247 voting more likely, but it also solidifies the majority advantage in other neighboring districts. It eliminates the possibility that minority voters could be pivotal swing voters and decreases the incentives of majority-group candidates to campaign to anyone other than their own coethnic supporters. In the short-term, it suggests that the way to compete against majority group ethnic parties is not by countering with minority group ethnic parties. Instead, broadly universal appeals which promise benefits to all voters regardless of ethnicity are most likely to appeal both to ethnic minority groups voters who have strong reasons to disapprove of ethnic parties from the majority group, and voters of the majority groups who may ultimately be unpersuaded by ethnic appeals from candidates in their own group. Given the numerical advantage that majority groups have by definition, they can only be defeated by multi-ethnic coalitions, and non-ethnic platforms and mobilization efforts are the most likely to be successful in that vein. The results also suggest that for those trying to defeat ethnic parties and candidates running on ethnic platforms which appeal to majority groups, the inclusion of ethnic minorities is vitally important. Ethnic minority voters and leaders are those with the strongest interests in preventing the politicization of ethnic identities, and given the non-viability of their own ethnic party a non-ethnic party is likely their best option. Many of the most successful ethnic minority political entrepreneurs are those who cannily adopted this strategy of building multi-ethnic movements based on universal values. Martin Luther King Jr. spoke consistently of justice, equality, and faith, and in doing so attracted support from white students and middle-class voters who may have been put off by more extreme black nationalism movements. César Chávez began his career by mobilizing Latino laborers, but ultimately founded an organization that represented all agricultural workers by recruiting Filipino, black, East Asian, and white blue-collar workers into the United Farm Workers labor union. Leaders such as these are rightly hailed as important people who won important social and political changes for minority groups, but it should not be forgotten that they also were leaders in building coalitions of interest with those outside their own ethnic community. Democracy may very well be the best regime type available for the protection of individual civil liberties. But it should not be forgotten that it is ultimately a system of majority rule, and

248 can, under certain conditions, support the arbitrary division of people and the exploitation of one group over another. Ethnic mobilization can undercut the democratic ideals of representation and accountability. By drawing attention to the fundamentally diﬀerent ways in which minority and majority group members are represented by democratic institutions, it is my hope that this dissertation increases our understanding of how democracy can function eﬀectively in divided and diverse societies.

249 Bibliography

Agarin, Timofey. 2010. A Cat’s Lick: Democratisation and Minority Communities in the Post- Soviet Baltic. Amsterdam: Rodopi.

Agarin, Timofey. 2011. “Civil Society versus Nationalizing State? Advocacy of Minority Rights in the Post-socialist Baltic states.” Nationalities Papers 39(2):181–203.

Albertson, Bethany L. 2015. “Dog-whistle Politics: Multivocal Communication and Religious Appeals.” Political Behavior 37(1):3–26.

Albrecht, Catherine. 1987. “Environmental Policies and Politics in Contemporary Czechoslo- vakia.” Studies in Comparative Communism 20(3-4):291–302.

Aldrich, John H. 1993. “Rational Choice and Turnout.” American Journal of Political Science pp. 246–278.

Alesina, Alberto, Reza Baqir & William Easterly. 1999. “Public Goods and Ethnic Divisions.” The Quarterly Journal of Economics 114(4):1243–1284.

Allcock, John B. 2000. Explaining Yugoslavia. New York: Columbia University Press.

Allen, Natalie J & John P Meyer. 1990. “The Measurement and Antecedents of Aﬀective, Continuance and Normative Commitment to the Organization.” Journal of Occupational and Organizational Psychology 63(1):1–18.

Alonso, Sonia & Rubén Ruiz-Ruﬁno. 2007. “Political Representation and Ethnic Conﬂict in new Democracies.” European Journal of Political Research 46(2):237–267.

Armakolas, Ioannis & Maja Maksimovic. 2014. “Memory and the Uses of the Wartime past in Contemporary Bosnia and Herzegovina: The Case of the Bosniak Campaign for the October 2013 Population Census.” Science and Society 32:59–93.

Asal, Victor, Amy Pate & Jonathan Wilkenfeld. 2008. “Minorities at Risk Organizational Be- havior Data and Codebook Version 9/2008 Online.”. URL: http://www.mar.umd.edu/data.asp

Aspinall, Edward. 2011. “Democratization and Ethnic Politics in Indonesia: Nine Theses.” Journal of East Asian Studies 11(2):289–319.

Auers, Daunis. 2012. “The curious case of the Latvian Greens.” Environmental Politics 21(3):522–527.

250 Aura. 2018. “Svetac kojeg poštuju židovi, kršćani i muslimani: Do podne Ilija, popodne Alija.” Dnevni avaz . Accessed: October 7, 2018. URL: "https://avaz.ba/zabava/zanimljivosti/283262/svetac-kojeg-postuju-zidovi-krscani-i- muslimani-do-podne-ilija-popodne-alija"

Bahtić-Kunrath, Birgit. 2011. “Of Veto Players and Entity-voting: Institutional Gridlock in the Bosnian Reform Process.” Nationalities Papers 39(6):899–923.

Bailey, Kenneth D. 1994. Typologies and Taxonomies: an Introduction to Classiﬁcation Tech- niques. Thousand Oaks: Sage Publications.

Bakker, Ryan, Catherine De Vries, Erica Edwards, Liesbet Hooghe, Seth Jolly, Gary Marks, Jonathan Polk, Jan Rovny, Marco Steenbergen & Milada Anna Vachudova. 2015. “Measuring Party Positions in Europe: The Chapel Hill Expert Survey Trend File, 1999–2010.” Party Politics 21(1):143–152.

Barbour, Henry, Sally Bradshaw, Ari Fleischer & Zori Fonalledas. 2013. “Growth and Opportunity Project.”. Accessed: October 31, 2019. URL: http://s3.documentcloud.org/documents/623664/republican-national-committees- growth-and.pdf

Barbullushi, Odeta. 2016. “The Limits and Ambiguities of the Albanian “National Question” in Post-communism: Political Parties, Albanian Nationalism, and External Actors.” East Eu- ropean Politics and Societies 30(3):551–570.

Barreto, Matt A, Gary M Segura & Nathan D Woods. 2004. “The Mobilizing Eﬀect of Majority– Minority Districts on Latino Turnout.” American Political Science Review 98(1):65–75.

Barth, Fredrik. 1969. Ethnic Groups and Boundaries: The Social Organization Of Culture Diﬀerence. Boston: Little, Brown.

Bates, Robert H, Avner Greif, Margaret Levi, Jean-Laurent Rosenthal & Barry R Weingast. 1998. Analytic Narratives. Princeton: Princeton University Press.

Beekman, Daniel. 2012. “Albanian-American realtor challenges incumbent Bronx Assembly- woman.” New York Daily News . April 23, 2012. Accessed: February 22, 2018. URL: http://www.nydailynews.com/new-york/bronx/realtor-albanian-american-elected- bronx-incumbent-experience-party-support-article-1.1065095

Belloni, Roberto. 2004. “Peacebuilding and Consociational Electoral Engineering in Bosnia and Herzegovina.” International Peacekeeping 11(2):334–353.

Bennich-Bjorkman, Li. 2017. "You are not the people": Revisiting Citizenship and Geopolitics. In Latvia–a Work in Progress? 100 Years of State- and Nation-building, ed. David J. Smith. Number vol. 142 in “Soviet and Post-Soviet Politics and Society” Stuttgart: Ibidem-Verlag pp. 175–188.

Bennich-Björkman, Li & Karl Magnus Johansson. 2012. “Explaining Moderation in National- ism: Divergent Trajectories of National Conservative Parties in Estonia and Latvia.” Compar- ative European Politics 10(5):585–607.

Benoit, Kenneth & Michael Laver. 2006. Party Policy in Modern Democracies. Routledge.

251 Bentley, G Carter. 1987. “Ethnicity and Practice.” Comparative Studies in Society and History 29(1):24–55.

Beriker-Atiyas, Nimet & Tijen Demirel-Pegg. 2000. “An Analysis of Integrative Outcomes in the Dayton Peace Negotiations.” International Journal of Conﬂict Management; Bowling Green 11(4):358–377.

Bershidsky, Leonid. 2018. “Putin Shouldn’t Feel Too Good About Russians in the Baltics.” Bloomberg.com . October 10, 2018. Accessed: July 19, 2019. URL: https://www.bloomberg.com/opinion/articles/2018-10-10/putin-shouldn-t-feel-too- good-about-pro-russian-triumph-in-latvia

Bhabha, Homi K. 1994. The Location of Culture. London: Routledge.

Bhaumik, Sumon Kumar, Ira N Gang & Myeong-Su Yun. 2006. “Ethnic conﬂict and Economic Disparity: Serbians and Albanians in Kosovo.” Journal of Comparative Economics 34(4):754– 773.

Bieber, Florian. 2005. “Local Institutional Engineering: A Tale of Two Cities, Mostar and Brčko.” International Peacekeeping 12(3):420–433.

Bieber, Florian. 2013. “When Counting Counts. The Bosnian Census.”. October 16, 2013. Accessed: October 3, 2018. URL: "https://ﬂorianbieber.org/2013/10/16/when-counting-counts-the-bosnian-census/"

Bilefsky, Dan. 2008. “Fears of New Ethnic Conﬂict in Bosnia.” The New York Times . December 14, 2008. Accessed: November 5, 2018. URL: https://www.nytimes.com/2008/12/14/world/europe/14bosnia.html

Bilefsky, Dan. 2013. “Protest in Bosnia Over IDs Traps Hundreds in Parliament.” The New York Times p. A4. June 8, 2013. Accessed: September 26, 2018. URL: https://www.nytimes.com/2013/06/08/world/europe/protest-in-bosnia-over-ids-traps- hundreds-in-parliament.html

Bird, Karen. 2014. “Ethnic Quotas and Ethnic Representation Worldwide.” International Polit- ical Science Review 35(1):12–26.

Birnir, Jóhanna Kristín. 2007. Ethnicity and Electoral Politics. Cambridge: Cambridge Univer- sity Press,.

Blake, Aaron. 2016. “Three Dozen Republicans have now Called for Donald Trump to Drop Out.” The Washington Post . October 9, 2016. Accessed: October 29, 2019. URL: https://www.washingtonpost.com/news/the-ﬁx/wp/2016/10/07/the-gops-brutal- responses-to-the-new-trump-video-broken-down/

BNN. 2012. “What is the History Behind Latvia’s National Flag?” Baltic News Network . November 16, 2012. Accessed: June 18, 2019. URL: https://bnn-news.com/history-latvias-national-ﬂag-81494

Bochsler, Daniel. 2011. “It is not How Many Votes You Get, but also where You Get Them. Territorial Determinants and Institutional Hurdles for the SUCCESs of Ethnic Minority Parties in Post-communist Countries.” Acta Politica 46(3):217–238.

252 Bogaards, Matthijs, Matthias Basedau & Christof Hartmann. 2010. “Ethnic Party Bans in Africa: an Introduction.” Democratization 17(4):599–617.

Bonilla-Silva, Eduardo. 2006. Racism without Racists: Color-Blind Racism and the Persistence of Racial Inequality in the United States. Lanham: Rowman & Littleﬁeld Publishers.

Bormann, Nils-Christian & Matt Golder. 2013. “Democratic Electoral Systems around the World, 1946–2011.” Electoral Studies 32(2):360–369.

Botev, Nikolai. 1994. “Where East Meets West: Ethnic Intermarriage in the Former Yugoslavia, 1962 to 1989.” American Sociological Review 59(3):461–480.

Bothwell, Laura E., Jeremy A. Greene, Scott H. Podolsky & David S. Jones. 2016. “Assess- ing the Gold Standard — Lessons from the History of RCTs.” The New England Journal of Medicine 374(22):2175–2181.

BPS - Sefer Halilović. 2016. “Program.” http://bps.ba/wp-content/themes/bps/pdf/ ProgramBPS-a.pdf.

Brader, Ted & Joshua A Tucker. 2001. “The Emergence of Mass Partisanship in Russia, 1993- 1996.” American Journal of Political Science pp. 69–83.

Bradt, Hale. 2004. Astronomy Methods: A Physical Approach to Astronomical Observations. Cambridge: Cambridge University Press.

Brancati, Dawn. 2006. “Decentralization: Fueling the Fire or Dampening the Flames of Ethnic Conﬂict and Secessionism?” International Organization 60(3):651–685.

Bratton, Michael & Mwangi S Kimenyi. 2008. “Voting in Kenya: Putting ethnicity in perspective.” Journal of Eastern African Studies 2(2):272–289.

Bratton, Michael, Robert Mattes & Emmanuel Gyimah-Boadi. 2005. Public Opinion, Democ- racy, and Market Reform in Africa. New York: Cambridge University Press.

Brewer, Marilynn B. 1991. “The Social Self: On Being the Same and Diﬀerent at the Same Time.” Personality and Social Psychology Bulletin 17(5):475–482.

Bringa, Tone. 1991. Being Muslim the Bosnian Way: Identity and Community in a Central Bosnian Village. Princeton: Princeton University Press.

Bringa, Tone R. 1993. “Nationality Categories, National Identiﬁcation AND Identity Formation In “Multinational” Bosnia.” Anthropology of East Europe Review 11(1 & 2):80–89.

Brubaker, Rogers. 1995. “National Minorities, Nationalizing States, and External National Homelands in the New Europe.” Daedalus 124(2):107–132.

Brubaker, Rogers. 2009. “Ethnicity, Race, and Nationalism.” Annual Review of Sociology 35:21–42.

Brubaker, Rogers et al. 1999. “The Manichean Myth: Rethinking the Distinction Between “Civic” and “Ethnic” Nationalism.” Nation and National Identity: The European Experience In Perspective pp. 55–71.

253 Brubaker, W Rogers. 1992. “Citizenship Struggles in Soviet Successor States.” International Migration Review 26(2):269–291.

Budge, Ian. 2003. Validating the Manifesto Research Group Approach: Theoretical Assump- tions and Empirical Conﬁrmations. In Estimating the Policy Position of Political Actors. Rout- ledge pp. 70–85. Bugajski, Janusz. 2002. Political Parties of Eastern Europe: a Guide to Politics in the Post- communist Era. Armonk: ME Sharpe.

Burden, Barry C. 2009. “Candidate-driven Ticket Splitting in the 2000 Japanese Elections.” Electoral Studies 28(1):33–40.

Burden, Barry C & David C Kimball. 2009. Why Americans Split their Tickets: Campaigns, Competition, and Divided Government. Ann Arbor: University of Michigan Press.

Burg, Steven L. & Paul Shoup. 1999. The War in Bosnia-Herzegovina : Ethnic Conﬂict and International Intervention. Armonk, N.Y.: M.E. Sharpe.

Burns, Alexander. 2015. “Choice Words From Donald Trump, Presidential Candidate.” The New York Times . June 16, 2015. Accessed: October 31, 2019. URL: https://www.nytimes.com/politics/ﬁrst-draft/2015/06/16/choice-words-from-donald- trump-presidential-candidate/

Cameron, Charles, David Epstein & Sharyn O’Halloran. 1996. “Do Majority-Minority Districts Maximize Substantive Black Representation in Congress?” American Political Science Review 90(4):794–812. Campbell, Angus, Philip E. Converse, Warren E. Miller & Donald E. Stokes. 1980[1960]. The American Voter. Chicago: University of Chicago Press.

Carby, Hazel V. 1992. “The Multicultural Wars.” Radical History Review 1992(54):7–18.

Carmichael, Cathie. 2015. A Concise History of Bosnia. New York: Cambridge University Press. Carney, Nikita. 2016. “All Lives Matter, but so does Race: Black Lives Matter and the Evolving Role of Social Media.” Humanity & Society 40(2):180–199.

Cebotari, Victor & Maarten P Vink. 2013. “A Conﬁgurational Analysis of Ethnic Protest in Europe.” International Journal of Comparative Sociology 54(4):298–324.

Cederman, Lars-Erik, Andreas Wimmer & Brian Min. 2010. “Why do Ethnic Groups Rebel? New Data and Analysis.” World Politics 62(1):87–119.

Central Intelligence Agency. 2017. The World Factbook. Accessed: January 6, 2018. URL: https://www.cia.gov/library/publications/the-world-factbook/

Central Statistical Bureau of Latvia. 2011. “Population Census.”. Accessed: May 14, 2019. URL: https://www.csb.gov.lv/en/statistics/statistics-by-theme/population/census

CERN. 2014. “New Results Indicate that New Particle is a Higgs Boson.” European Organiza- tion for Nuclear Research. March 14, 2014. Accessed: February 26, 2019. URL: https://home.cern/news/news/physics/new-results-indicate-new-particle-higgs-boson

254 G, ermanis,¯ Uldis. 2007. The Latvian Saga. Riga: Atena.¯ Chandra, Kanchan. 2004. Why Ethnic Parties Succeed: Patronage and Ethnic Head Counts in India. Cambridge: Cambridge University Press.

Chandra, Kanchan. 2011. “What is an Ethnic Party?” Party Politics 17(2):151–169.

Chandra, Kanchan. 2012a. Constructivist Theories of Ethnic Politics. Oxford University Press. Chandra, Kanchan, ed. 2012b. Constructivist Theories of Ethnic Politics. New York: Oxford University Press.

Chang, Eric CC, Miriam A Golden & Seth J Hill. 2010. “Legislative Malfeasance and Political Accountability.” World Politics 62(2):177–220. Charter, David. 2013. “Death of a Language: Last Ever Speaker of Livonian Passes Away Aged 103.” The Times . June 5, 2013. Accessed: August 22, 2017. URL: https://www.thetimes.co.uk/article/death-of-a-language-last-ever-speaker-of-livonian- passes-away-aged-103-8k0rlplv8xj

Chinn, Jeﬀ & Lise A Truex. 1996. “The Question of Citizenship in the Baltics.” Journal of Democracy 7(1):133–147. Cho, Wendy K Tam. 1998. “Iﬀ the Assumption Fits...: A Comment on the King Ecological Inference Solution.” Political Analysis 7:143–163.

Christia, Fotini. 2008. “Following the Money: Muslim versus Muslim in Bosnia’s Civil War.” Comparative Politics 40(4):461–480.

Clark, Terry D. 2006. Nationalism in Post-Soviet Lithuania: New Approaches for the Nation of “ Innocent Suﬀerers”. In After Independence: Making and Protecting the Nation in Postcolonial and Postcommunist States, ed. Lowell W. Barrington. Ann Arbor: University of Michigan Press pp. 162–186.

Clark, William Roberts & Matt Golder. 2006. “Rehabilitating Duverger’s Theory: Testing the Mechanical and Strategic Modifying Eﬀects of Electoral Laws.” Comparative Political Studies 39(6):679–708.

Clemens, Walter C. 1991. Baltic Independence and Russian Empire. New York: St. Martin’s Press.

Collier, David & Colin Elman. 2008. Qualitative and Multimethod Research: Organizations, Publications, and Reﬂection on Integration. In The Oxford Handbook of Political Methodology, ed. Janet M. Box-Steﬀensmeier & Henry E. Brady. The Oxford Handbooks of Political Science Oxford ; New York: Oxford University Press.

Collier, David, Jody LaPorte & Jason Seawright. 2012. “Putting Typologies to Work: Concept Formation, Measurement, and Analytic Rigor.” Political Research Quarterly 65(1):217–232.

Commercio, Michele E. 2008. “Systems of Partial Control: Ethnic Dynamics in Post-Soviet Estonia and Latvia.” Studies in Comparative International Development 43(1):81–100.

Converse, Philip E. 1969. “Of Time and Partisan Stability.” Comparative Political Studies 2(2):139–171.

255 Cooley, Charles Horton. 1922. Human Nature and the Social Order. New York: C. Scribner’s Sons.

Cornell, Stephen. 1996. “The Variable Ties that Bind: Content and Circumstance in Ethnic Processes.” Ethnic and Racial Studies 19(2):265–289.

Cox, Gary W. 1984. “Strategic Electoral Choice in Multi-Member Districts: Approval Voting in Practice?” American Journal of Political Science pp. 722–738.

Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World’s Electoral System. New York: Cambridge University Press.

Crenshaw, Kimberlé. 1991. “Mapping the Margins: Intersectionality, Identity Politics, and Violence Against Women Of Color.” Stanford law review pp. 1241–1299.

Csergő, Zsuzsa. 2007. Talk of the Nation: Language and Conﬂict in Romania and Slovakia. Ithaca: Cornell University Press.

Dalton, Harlon. 2008. Failing to See. In White Privilege: Essential Readings on the Other Side of Racism, ed. Paula S. Rothenberg. New York: Worth pp. 15–18.

Dancygier, Rafaela M. 2017. Dilemmas of Inclusion: Muslims in European politics. Princeton: Princeton University Press.

Davies, Philip John & Andrejs Valdis Ozolins. 1994. “The Latvian Parliamentary Election of 1993.” Electoral Studies 13(1):83–86.

Dawsey, Josh. 2018. “Trump Derides Protections for Immigrants from ‘Shithole’ Countries.” The Washington Post . January 12, 2018. Accessed: October 29, 2019. URL: https://www.washingtonpost.com/politics/trump-attacks-protections-for-immigrants- from-shithole-countries-in-oval-oﬃce-meeting/2018/01/11/bfc0725c-f711-11e7-91af- 31ac729add94_story.html

Deets, Stephen & Benjamin McClelland. 2018. “Europeanization in Bosnia and Herzegovina: Recent Lessons from the Dairy Sector.” Problems of Post-Communism pp. 1–14. URL: https://www.tandfonline.com/doi/full/10.1080/10758216.2018.1484666

DELFI.LV. 2012. “Ушаков: лучше сдать на гражданство, чем сидеть и жаловаться.” DELFI.LV . May 29, 2012. Accessed: July 19, 2019. URL: https://rus.delﬁ.lv/news/daily/latvia/ushakov-luchshe-sdat-na-grazhdanstvo-chem- sidet-i-zhalovatsya.d?id=42391186&all=true

Demokratska Fronta. 2014. “Demokratska Fronta - Manifest.” http://www.demokratskafronta. ba/index.php?lang=ba&sel=28.

Denzin, Norman K. 2009. The Research Act : a Theoretical Introduction to Sociological Meth- ods. New Brunswick, NJ: AldineTransaction.

DeVotta, Neil. 2004. Blowback: Linguistic nationalism, Institutional Decay, and Ethnic Conﬂict in Sri Lanka. Palo Alto: Stanford University Press.

Dhamoon, Rita Kaur. 2011. “Considerations on Mainstreaming Intersectionality.” Political Re- search Quarterly 64(1):230–243.

256 Dimitrovova, Bohdana. 2001. “Bosniak or Muslim? Dilemma of one Nation with two Names.” Southeast European Politics 2(2):94–108.

Dnevni avaz. 2011. “SNSD suspendovan iz Socijalističke internacionale.” http://archive.is/ A26w.

DNZ BiH. 2008. “O nama.”. URL: http://dnzbih.ba/kategorije/o_nama

Donia, Robert J. 2000. “Review Article: the New Bosniak History.” Nationalities Papers 28(2):351–358.

Downs, Anthony. 1957. An Economic Theory of Democracy. NewYork: Harper and Bros.

Dreifelds, Juris. 1996. Latvia in Transition. Cambridge: Cambridge University Press.

Dunning, Thad & Janhavi Nilekani. 2013. “Ethnic Quotas and Political Mobilization: Caste, Parties, and Distribution in Indian village Councils.” American Political Science Review 107(1):35–56.

Dunning, Thad & Lauren Harrison. 2010. “Cross-cutting Cleavages and Ethnic Voting: An Experimental Study of Cousinage in Mali.” American Political Science Review 104(1):21–39.

Duverger, Maurice. 1954. Political Parties: Their Organization and Activity in the Modern State. New York: Wiley.

Dvornik, Francis. 1962. The Slavs in European History and Civilization. New Brunswick, N.J.: Rutgers University Press.

Dyer, Richard. 1997. The Matter of Whiteness. In White: Essays on Race and Culture, ed. Richard Dyer. pp. 1–40.

Džankić, Jelena. 2015. “The Politics of Inclusion and Exclusion: Citizenship and Voting Rights in Bosnia and Herzegovina.” International Peacekeeping 22(5):526–544.

Eakins, Emily. 2016. “Policing in America: Understanding Public Attitudes Toward the Police. Results from a National Survey.”. Accessed: October 31, 2019. URL: https://www.cato.org/survey-reports/policing-america

Easley, Jonathan. 2017. “Poll: 57 Percent Have Negative View of Black Lives Matter movement.” The Hill . August 2, 2017. Accessed: October 31, 2019. URL: https://thehill.com/homenews/campaign/344985-poll-57-percent-have-negative- view-of-black-lives-matter-movement

Easterly, William & Ross Levine. 1997. “Africa’s Growth Tragedy: Policies and Ethnic Divi- sions.” The Quarterly Journal of Economics 112(4):1203–1250.

Eglitis, Daina S & Laura Ardava. 2012. “The Politics of Memory: Remembering the Baltic Way 20 Years after 1989.” Europe-Asia Studies 64(6):1033–1059.

Eglitis, Daina Stukuls. 2002. Imagining the Nation: History, Modernity, and Revolution in Latvia. University Park, Pa: Pennsylvania State University Press.

257 ‘ Egl¯ıtis, Ol,gerts. 1993. Nonviolent Action in the Liberation of Latvia. Cambridge, MA: Albert Einstein Institution.

Eifert, Benn, Edward Miguel & Daniel N Posner. 2010. “Political Competition and Ethnic Identiﬁcation in Africa.” American Journal of Political Science 54(2):494–510.

European Election Database. 2018. “Latvia - Political parties.”. Accessed: May 8, 2019. URL: https://nsd.no/european_election_database/country/latvia/parties.html

Evans, Geoﬀrey & Stephen Whiteﬁeld. 1998. “The Evolution of Left And Right in Post-Soviet Russia.” Europe-Asia Studies 50(6):1023–1042.

Fagin, Adam. 1994. “Environment and Transition in the Czech Republic.” Environmental Pol- itics 3(3):479–494.

Fearon, J. D. 1999a. Why Ethnic Politics and "Pork" Tend to go Together. In Ethnic Politics and Democratic Stability. University of Chicago: Politics and Democratic Stability.

Fearon, James D. 1999b. “What Is Identity (As We Now Use the Word)?” https://web.stanford.edu/group/fearon-research/cgi-bin/wordpress/wp-content/uploads/ 2013/10/What-is-Identity-as-we-now-use-the-word-.pdf.

Fearon, James D & David D Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American political science review 97(1):75–90.

Ferree, Karen E. 2006. “Explaining South Africa’s Racial Census.” The Journal of Politics 68(4):803–815.

Fiske, Susan T & Shelley E Taylor. 2013[1984]. Social Cognition: From Brains to Culture. Los Angeles: Sage.

Flores, Morse Caoagas. 2008. “The Role of International Labour Organization in the Promotion and Protection of Indigenous Languages.”. URL: http://www.un.org/esa/socdev/unpﬁi/documents/EGM_IL_Flores.doc

Fraenkel, Jon & Bernard Grofman. 2006. “Does the Alternative Vote Foster Moderation in Ethnically Divided Societies? The Case of Fiji.” Comparative Political Studies 39(5):623–651.

France 24. 2015. “Latvia elects EU’s ﬁrst Green president.”. June 4, 2015. Accessed June 18, 2019. URL: https://www.france24.com/en/20150604-latvia-eu-ﬁrst-green-president-vejonis

Freedman, David A. 2008. On Types of Scientiﬁc Enquiry: the Role of Qualitative Reasoning. In The Oxford Handbook of Political Methodology, ed. Janet M Box-Steﬀensmeier, Henry E Brady & David Collier. Vol. 10 New York: Oxford University Press pp. 300–318.

Freedman, David A, Stephen P Klein, Jerome Sacks, Charles A Smyth & Charles G Everett. 1991. “Ecological Regression and Voting Rights.” Evaluation Review 15(6):673–711.

Frejut˙ e-Rakauskien˙ e,˙ Monika. 2011. “The Impact of the EU membership on Ethnic Minority Participation. Parties of Lithuanian Ethnic Minorities in the European Parliament Elections.” Politikos mokslu˛almanachas (10):7–30.

258 Galbreath, David J. & Daunis Auers. 2010. Green, Black, and Brown: Uncovering Latvia’s En- vironmental Politics. In Contemporary environmentalism in the Baltic States : from phosphate springs to ’Nordstream’, ed. David J Galbreath. London: Routledge pp. 54–59.

Geertz, Cliﬀord. 1963. “Old Societies and New States: The Quest for Modernity in Asia and Africa.”.

Gellner, Ernest. 1983. Nations and Nationalism. Oxford, England: Blackwell.

Gherghina, Sergiu & George Jiglau. 2011. “Explaining Ethnic Mobilisation in Post-communist Countries.” Europe-Asia Studies 63(1):49–76.

Gil-White. 2001. “Are Ethnic Groups Biological "Species" to the Human Brain? Essentialism in Our Cognition of Some Social Categories.” Current Anthropology 42(4):515.

Ginkel, John. 2002. “Identity Construction in Latvia’s “Singing Revolution”: Why Inter-ethnic Conﬂict Failed to Occur.” Nationalities Papers 30(3):403–433.

Givens, Terri E & Rahsaan Maxwell. 2012. Immigrant Politics: Race and Representation in Western Europe. Boulder: Lynne Rienner.

Golder, Sona Nadenichek. 2006. “Pre-electoral Coalition Formation in Parliamentary Democ- racies.” British Journal of Political Science 36(2):193–212.

Goldstein, Ellen & Fengler Davies, Simon and Wolfgang. 2015. “Three Reasons why the Economy of Bosnia and Herzegovina is oﬀ Balance.”. URL: https://www.brookings.edu/blog/future-development/2015/11/05/three-reasons-why- the-economy-of-bosnia-and-herzegovina-is-oﬀ-balance/

Goldstein, Lewis H. 2015. “Assemblyman Mark Gjonaj Receives Albania’s Highest Medal of Honor.” The Bronx Chronicle . May 9, 2015. Accessed February 2, 2018. URL: http://thebronxchronicle.com/2015/05/08/assemblyman-mark-gjonaj-receives- albanias-highest-medal-honor/

Goode, J Paul & David R Stroup. 2015. “Everyday Nationalism: Constructivism for the Masses.” Social Science Quarterly 96(3):717–739.

Gosnell, Harold F. 1957. Machine Politics: Chicago model. Chicago: University of Chicago Press.

Grofman, Bernard & Chandler Davidson, eds. 1992. Controversies in Minority Voting: The Voting Rights Act in Perspective. Washington, D.C: Brookings Institution.

Grzymala-Busse, Anna M. 2002. Redeeming the Communist Past: The Regeneration of Com- munist Parties in East Central Europe. Cambridge University Press.

Gurr, Ted Robert. 1970. Why Men Rebel. Princeton, NJ: Princeton University Press.

Gurr, Ted Robert. 2000. Peoples Versus States: Minorities at Risk in the New Century. Wash- ington: US Institute of Peace Press.

HadžiMuhamedović, Safet. 2018. Waiting for Elijah : Time and Encounter in a Bosnian land- scape. New York: Berghahn Books.

259 Hale, Henry E. 2004. “Explaining Ethnicity.” Comparative Political Studies 37(4):458–485.

Hall, Brian. 1994. Impossible Country. Secker & Warbury.

Hall of Infamy: Milorad Dodik. 2019. New Internationalist (518):72–72.

Hancock, Ange-Marie. 2007. “Intersectionality as a Normative and Empirical Paradigm.” Pol- itics & Gender 3(2):248–254. Handelman, Don. 1977. “The Organization of Ethnicity.” Ethnic Groups 1(3):187–200.

Haney-López, Ian. 2015. Dog Whistle Politics: How Coded Racial Appeals Have Reinvented Racism and Wrecked the Middle Class. New York: Oxford University Press.

Harris-Perry, Melissa. 2016. “Did Gender Matter?” The Politics of Gender: Women, Men, and the 2016 Campaign. An Atlantic Politics & Policy Brieﬁng. December 13, 2016. Accessed: October 29, 2019. URL: http://www.federalreserve.gov/boarddocs/speeches/1996/19961205.htm

HDZ. 2007. Program Hrvatske demokratske zajednice Bosne i Hercegovine. Mostar: May 12, 2007. Heath, Anthony F, Stephen D Fisher, David Sanders & Maria Sobolewska. 2011. “Ethnic Het- erogeneity in the Social Bases of Voting at the 2010 British General Election.” Journal of Elections, Public Opinion and Parties 21(2):255–277.

Helsinki Declaration. 1975. “The Final Act of the Conference on Security and Cooperation in Europe, 14 I.L.M. 1292.”. Hill, Kevin A, Dario V Moreno & Lourdes Cue. 2001. “Racial and Partisan Voting in a Tri- Ethnic City: The 1996 Dade County Mayoral Election.” Journal of Urban Aﬀairs 23(3-4):291– 307.

Hillion, Christophe. 2004. The Copenhagen Criteria and their Progeny. In EU enlargement: a legal approach, ed. Christophe Hillion. Oxford: Hart pp. 1–22. OCLC: ocm52531038. Hirschfeld, Lawrence A. 1998. Race in the Making: Cognition, Culture, and the Child’s Con- struction of Human Kinds. Cambridge, MA: MIT Press.

Hodson, Randy, Dusko Sekulic & Garth Massey. 1994. “National Tolerance in the Former Yugoslavia.” American Journal of Sociology 99(6):1534–1558.

Holmes, Stephen. 1996. Cultural Legacies or State Collapse? Probing the Postcommunist Dilemma. In Postcommunism: four perspectives, ed. Michael Mandelbaum. New York: Coun- cil on Foreign Relations pp. 22–76.

Hooghe, Liesbet, Ryan Bakker, Anna Brigevich, Catherine De Vries, Erica Edwards, Gary Marks, Jan Rovny, Marco Steenbergen & Milada Vachudova. 2010. “Reliability and Valid- ity of the 2002 and 2006 Chapel Hill Expert Surveys on Party Positioning.” European Journal of Political Research 49(5):687–703. Hope, Nicholas. 1999. Interwar Statehood: Symbol and Reality. In The Baltic States: the Na- tional Self-determination of Estonia, Latvia and Lithuania, ed. Graham Smith. 1. published, digital printing ed. Basingstoke, Hampshire: Macmillan pp. 41–68. OCLC: 248347318.

260 Horowitz, Donald L. 1985. Ethnic Groups in Conﬂict. Berkeley: University of California Press.

Howard, Lise Morjé. 2012. “The Ethnocracy Trap.” Journal of Democracy 23(4):155–169.

How Donald Trump learned to love Confederate monuments. 2017. August 22, 2017. Accessed October 31, 2019. URL: https://www.economist.com/graphic-detail/2017/08/22/how-donald-trump-learned- to-love-confederate-monuments

Hromadžić, Azra. 2013. “Discourses of Trans-ethnic Narod in Postwar Bosnia and Herzegov- ina.” Nationalities Papers 41(2):259–275.

HSP BiH. 2015. “Deklaracija Hrvatske stranke prava Bosne i Hercegovine.”. Accessed: Novem- ber 11, 2018. URL: https://hsp-bih.ba/deklaracija-hrvatske-stranke-prava-bosne-i-hercegovine

Huber, John D. 2012. “Measuring Ethnic Voting: Do Proportional Electoral Laws Politicize Ethnicity?” American Journal of Political Science 56(4):986–1001.

Huber, John D. 2017. Exclusion by Elections: Inequality, Ethnic Identity, and Democracy. New York: Cambridge University Press.

Huber, John D & Pavithra Suryanarayan. 2016. “Ethnic Inequality and the Ethniﬁcation of Political Parties: Evidence from India.” World Politics 68(1):149–188.

Hughes, James. 2005. “‘Exit’ in Deeply Divided Societies: Regimes of Discrimination in Es- tonia and Latvia and the Potential for Russophone Migration.” Journal of Common Market Studies 43(4):739–762.

Hulsey, John. 2015. “Electoral Accountability in Bosnia and Herzegovina under the Dayton Framework Agreement.” International Peacekeeping 22(5):511–525.

Humphreys, Macartan & Alan M Jacobs. 2015. “Mixing Methods: A Bayesian Approach.” American Political Science Review 109(4):653–673.

Hutchinson, John & Anthony D. Smith. 1996. Introduction. In Ethnicity, ed. John Hutchinson & Anthony D. Smith. Oxford Readers Oxford: Oxford University Press pp. 3–14.

Ijabs, Ivars. 2014. “Another Baltic Postcolonialism: Young Latvians, Baltic Germans, and the emergence of Latvian National Movement.” Nationalities Papers 42(1):88–107.

Ikstens, Janis.¯ 2007. Does Europe Matter? The EU and Latvia’s Political Parties. In The Euro- pean Union and Party Politics in Central and Eastern Europe. Springer pp. 86–106.

Ikstens, Janis.¯ 2017. Leader-Centered and Power-Hungry: Party Organizations in Latvia. In Organizational Structures of Political Parties in Central and Eastern European Coun- tries, ed. Katarzyna Sobolewska-Myślik, Beata Kosowska-Gąstoł & Piotr Borowiec. Kraków: Wydawnictwo Uniwersytetu Jagiellońskiego pp. 231–250.

International Criminal Tribunal for the former Yugoslavia. 2005. “Judgement in the Case the Prosecutor v. Sefer Halilović.” http://www.icty.org/x/cases/halilovic/tjug/en/051116_ halilovic_summary_en.pdf.

261 Ishiyama, John. 2012. “Explaining Ethnic Bloc Voting in Africa.” Democratization 19(4):761– 788.

Is Serbo-Croatian a language? 2017. The Economist . April 10, 2017. Accessed: October 4, 2018. URL: https://www.economist.com/the-economist-explains/2017/04/10/is-serbo-croatian-a- language

Iwanek, Krzysztof. 2016. “The Curious Stories of Indian Party Symbols.” The Diplomat . Ac- cessed: November 15, 2018. URL: "https://thediplomat.com/2016/11/the-curious-stories-of-indian-party-symbols/"

Jancar-Webster, Barbara. 1993. Environmental Action in Eastern Europe: Responses to Crisis. Armonk: ME Sharpe.

Jeﬀrey, Alex. 2006. “Building State Capacity in Post-Conﬂict Bosnia and Herzegovina: The Case of Brčko District.” Political Geography 25(2):203–227.

Jenkins, Richard. 1996. Social Identity. London: Routledge.

Jenkins, Richard. 2000. “Categorization: Identity, Social Process and Epistemology.” Current sociology 48(3):7–25.

Jubulis, Mark A. 2007. The Persistence of the Baltic Nations under Soviet Rule. In Nationalism in a Global Era: The Persistence of Nations, ed. Mitchell Young, Eric Zuelow & Andreas Sturm. New York: Routledge pp. 179–197.

Kalnins, Mara. 2015. Latvia : a Short History. New York: Oxford University Press.

Kamenska, Anhelita & Ilze Brands-Kehris. 2008. Combating Hate Crimes in Latvia: Legisla- tion and Police Practice. Technical report Latvian Centre for Human Rights Riga: . Accessed: June 18, 2019. URL: http://cilvektiesibas.org.lv/site/attachments/30/01/2012/Naida_noziegums_ENG_ Internetam.pdf

Kasekamp, Andres. 2006. Latvia. In World Fascism: A Historical Encyclopedia, ed. Cyprian Blamires & Paul Jackson. Santa Barbara, CA: ABC-CLIO pp. 375–6.

Keil, Soeren & Valery Perry. 2015. “Back to square one? An analysis of the 2014 General Elections in Bosnia and Herzegovina.” Electoral Studies 38:82–87.

Kendall, Brent. 2016. “Trump Says Judge’s Mexican Heritage Presents ‘Absolute Conﬂict’.” The Wall Street Journal . June 3, 2016. Accessed October 29, 2019. URL: https://www.wsj.com/articles/donald-trump-keeps-up-attacks-on-judge-gonzalo- curiel-1464911442

Khlakovskiy, Vitaliy. 2018. “Стабильность и пенсии от €400. Что лидеры ’Согласия’ рас- сказали в студии Delfi TV.” delfi.lv . October 4, 2018. Accessed July 19, 2019. URL: https://www.delfi.lv/a/50457419&all=true

King, Gary. 1997. A Solution to the Ecological Inference Problem. Princeton: Princeton Uni- versity Press.

262 King, Gary, Robert O Keohane & Sidney Verba. 1994. Designing Social Inquiry: Scientiﬁc Inference in Qualitative Research. Princeton: Princeton University Press.

Kirisci, Kemal & Gareth M Winrow. 2013. The Kurdish Question and Turkey: An Example of a Trans-state Ethnic Conﬂict. London: Routledge.

Kitschelt, Herbert. 1992. “The Formation of Party Systems in East Central Europe.” Politics & Society 20(1):7–50.

Kitschelt, Herbert. 2001. Divergent Paths of Postcommunist Democracies. In Political Parties and Democracy, ed. Larry Jay Diamond & Richard Gunther. Journal of Democracy Baltimore, MD: Johns Hopkins University Press.

Knežević, Anto. 2001. “Inhabitants of the Proud Bosnia: the Identity of the European Native Muslims.” Islamic Studies 40(1):133–149.

Knezevic, Gordana. 2016. “From Pillar Of Dayton Bosnia To Its Weakest Link – Dodik.” RadioFreeEurope/RadioLiberty . September 13, 2016. Accessed: November 19, 2018. URL: https://www.rferl.org/a/commentary-from-pillar-of-dayton-to-weakest-link/ 27984037.html

Koinova, Maria. 2013. “Four Types of Diaspora Mobilization: Albanian Diaspora Activism for Kosovo Independence in the US and the UK.” Foreign Policy Analysis 9(4):433–453.

Kolstø, Pål. 1995. Russians in the Former Soviet Republics. London: Hurst & Company.

Kreuzer, Marcus & Vello Pettai. 2003. “Patterns of Political Instability: Aﬃliation Patterns of Politicians and Voters in Post-communist Estonia, Latvia, and Lithuania.” Studies in Compar- ative International Development 38(2):76–98.

Kuczyńska-Zonik, Aleksandra. 2017. “Non-citizens in Latvia: Is it a Real Problem?” Sprawy Narodowościowe (49).

La Porta, Rafael, Florencio Lopez-de Silanes, Andrei Shleifer & Robert Vishny. 1999. “The Quality of Government.” The Journal of Law, Economics, and Organization 15(1):222–279. Laakso, Markku & Rein Taagepera. 1979. “"Eﬀective" Number of Parties A Measure with Application to West Europe.” Comparative Political Studies 12(1):3–27.

Lagasse, Paul. 2017. “Identity.” The Columbia Encyclopedia .

Laitin, David D. 1998. Identity in Formation: The Russian-speaking Populations in the Near Abroad. Ithaca: Cornell University Press. Laitin, David D. 2006. “Ethnography and/or Rational Choice.” Qualitative Methods 4(1):26–33.

Latvijas Valsts Prezidenta Majas¯ Lapa. 2017. “Latvijas valsts gerbonis.”.‘ Accessed: June 18, 2019. URL: https://web.archive.org/web/20171005004954/http://president.lv/pk/content/?cat_ id=7029

LegislationLine. 2006. “Election Law of Bosnia and Herzegovina.”. Accessed November 5, 2018. URL: https://www.legislationline.org/documents/action/popup/id/6244

263 Less, Timothy. 2016. “Dysfunction in the Balkans.” Foreign Aﬀairs . Accessed September 26, 2018. URL: https://www.foreignaﬀairs.com/articles/bosnia-herzegovina/2016-12-20/dysfunction- balkans

Li, Yuhui. 2018. “Electoral System Eﬀects Re-examined Using the Largest Vote Share Vari- able.” Democratization 25(1):58–77.

Lieven, Anatol. 1994. The Baltic Revolution: Estonia, Latvia, Lithuania and the Path to Inde- pendence. New Haven: Yale University Press.

Lijphart, Arend. 1977. Democracy in Plural Societies: a Comparative Exploration. New Haven: Yale University Press. Lijphart, Arend, ed. 1994. Electoral Systems and Party Systems: A Study of Twenty-Seven Democracies, 1945-1990. Oxford: Oxford University Press.

Linz, Juan J. & Alfred Stepan. 1996. Problems of Democratic Transition and Consolidation: Southern Europe, South America, and Post-Communist Europe. Baltimore: Johns Hopkins University Press. Lipset, Seymour Martin & Stein Rokkan, eds. 1967. Party Systems and Voter Alignments: Cross-National Perspectives. New York: Free Press.

Liu, Amy H. 2015. Standardizing Diversity: the Political Economy of Language Regimes. Philadelphia: University of Pennsylvania Press.

Loewen, James W & Bernard Grofman. 1989. “Recent Developments in Methods Used in Vote Dilution Litigation.” The Urban Lawyer 21:589.

Lublin, David. 2014. Minority Rules: Electoral Systems, Decentralization, and Ethnoregional Party Success. New York: Oxford University Press.

Lupia, Arthur, Mathew D McCubbins, Lupia Arthur et al. 1998. The Democratic Dilemma: Can Citizens Learn what They Need to Know? New York: Cambridge University Press. Madrid, Raúl L. 2012. The Rise of Ethnic Politics in Latin America. New York :: Cambridge University Press.

Mahoney, James. 2012. “The Logic of Process Tracing Tests in the Social Sciences.” Sociolog- ical Methods & Research 41(4):570–597.

Mair, Peter. 1998. Party System Change: Approaches and Interpretations. New York: Oxford University Press.

Manin, Bernard, Adam Przeworski & Susan Stokes. 1999. “Elections and Representation.” Democracy, accountability, and representation 2:29.

Maravall, José María. 2005. Accountability and the Survival of Governments. In The Oxford Handbook of Comparative Politics, ed. Carles Boix & Susan C. Stokes. New York: Oxford University Press.

Martinelli, César. 2006. “Would Rational Voters Acquire Costly Information?” Journal of Economic Theory 129(1):225–251.

264 Martyn-Hemphill, Richard. 2016a. “Estonia’s New Premier Comes From Party With Links to Russia.” The New York Times . November 21, 2016. Accessed: October 20, 2019. URL: https://www.nytimes.com/2016/11/21/world/europe/estonia-juri-ratas-center- party.html

Martyn-Hemphill, Richard. 2016b. “Latvia Wants to Ban Face Veils, for All 3 Women Who Wear Them.” The New York Times . April 19, 2016. Accessed September 26, 2018. URL: https://www.nytimes.com/2016/04/20/world/europe/latvia-face-veils-muslims- immigration.html

McGarry, John & Brendan O’Leary. 2006. “Consociational Theory, Northern Ireland’s Con- ﬂict, and its Agreement. Part 1: What Consociationalists can Learn from Northern Ireland.” Government and Opposition 41(1):43–63.

McKelvey, Richard D & Peter C. Ordeshook. 1972. A General Theory of the Calculus of Voting. In Mathematical Applications in Political Science. Charlottesville: University of Virginia Press pp. 32–78.

Mead, George Herbert. 1934. Mind, Self and Society. Chicago: Chicago University of Chicago Press.

Mendelberg, Tali. 2001. The Race Card: Campaign Strategy, Implicit Messages, and the Norm of Equality. Princeton: Princeton University Press.

Mikkel, Evald & Geoﬀrey Pridham. 2004. “Clinching the ‘Return to Europe’: The Referendums on EU Accession in Estonia and Latvia.” West European Politics 27(4):716–748.

Milekic, Sven. 2017. “Post-Yugoslav ‘Common Language’ Declaration Challenges National- ism.” Balkan Insight . March 29, 2017. Accessed: October 4, 2018. URL: "http://www.balkaninsight.com/en/article/post-yugoslav-common-language- declaration-challenges-nationalism-03-29-2017"

Misha, Glenny. 1996. The Fall of Yugoslavia: The Third Balkan War. New York: Penguin.

Misiunas, Romuald J & Rein Taagepera. 1993. The Baltic States, Years of Dependence, 1940- 1990. Berkeley: Univ of California Press.

Mitchell, Paul, Geoﬀrey Evans & Brendan O’Leary. 2009. “Extremist Outbidding in Eth- nic Party Systems is Not Inevitable: Tribune Parties in Northern Ireland.” Political Studies 57(2):397–421.

Montalvo, José G. & Marta Reynal-Querol. 2005. “Ethnic Polarization, Potential Conﬂict, and Civil Wars.” The American Economic Review 95(3):796–816.

Morlino, Leonardo. 1998. Democracy Between Consolidation and Crisis: Parties, Groups, and Citizens in Southern Europe. New York: Oxford University Press.

Morris, Helen M. 2003. “EU Enlargement and Latvian Citizenship Policy.” Journal on Ethnop- olitics and Minority Issues in Europe pp. 1–39.

Mudde, Cas. 2014. “EU Earthquake? Not Quite.”. June 2, 2014, Accessed: August 22, 2017. URL: https://www.hate-speech.org/of-earthquakes-cartels-and-and-designer-fascists/

265 Muižnieks, N¯ıls R. 1987. “The Daugavpils Hydro Station And "Glasnost" In Latvia.” Journal of Baltic Studies 18(1):63–70.

Muiznieks, Nils. 2011. “Max van der Stoel and Latvia.” Security & Human Rights 22(3):257– 260.

Mujanović, Jasmin. 2014. “Bosnia-Herzegovina Election Guide: Personalities, Parties & Prospects.”. September 24, 2014. Accessed: November 13, 2018. URL: https://balkanist.net/bosnia-herzegovina-election-guide-personalities-parties- prospects/

Mylonas, Christos. 2003. Serbian Orthodox Fundamentals : the Quest for an Eternal Identity. New York: Central European University Press.

Nakai, Ryo. 2014. “The Inﬂuence of Party Competition on Minority Politics: A Comparison of Latvia and Estonia.” Journal on Ethnopolitics and Minority Issues in Europe 13:57.

Naomi Rivera Defeated Soundly in Assembly Primary By Hard-Working Challenger. 2012. Norwood News . September 20 2012, Accessed: Februrary 23, 2018. URL: http://www.norwoodnews.org/id=9050&story=naomi-rivera-defeated-soundly-in- assembly-primary-by-hard-working-challenger/

Nardelli, Alberto, Denis Dzidic & Elvira Jukic. 2014. “Bosnia and Herzegovina: the World’s Most Complicated System of Government?” The Guardian .

Neto, Octavio Amorim & Gary W Cox. 1997. “Electoral Institutions, Cleavage Structures, and the Number of Parties.” American Journal of Political Science pp. 149–174.

Niškanović, Miroslav. 1978. “Ilindanski dernek kod turbeta Djerzelez Alije u Gerzovu.” Novi Pazar pp. 163–168.

Nordlinger, E. A. 1972. Conﬂict Regulation in Divided Societies. Cambridge: Harvard Univer- sity.

Novagrockiene,˙ Jurate.˙ 2001. “The Development and Consolidation of the Lithuanian Political Party System.” Journal of Baltic studies 32(2):141–155.

NSRzB. 2004. “Programska Načela Narodne stranke Radom za boljitak.”. URL: http://www.zaboljitak.ba/wp-content/uploads/2010/05/programska_nacela_bos.pdf

NSRzB. 2008. “Statut Narodne stranke Radom za boljitak.”. URL: http://www.zaboljitak.ba/index.php/dokumenti/statut-narodne-stranke-radom-zaboljitak/

Nugent, Paul. 2001. “Ethnicity as an Explanatory Factor in the Ghana 2000 Elections.” African Issues 29(1-2):2–7.

O’Ballance, Edgar. 2016. Civil War in Bosnia 1992–94. Springer.

Ordeshook, Peter C. & Olga V. Shvetsova. 1994. “Ethnic Heterogeneity, District Magnitude, and the Number of Parties.” American Journal of Political Science 38(1):100.

266 Oslobođenje. 2017. “Čović za HRT: Treći entitet je realna mogućnost, trebali smo ga tražiti prije 20 godina.”. Decmeber 17, 2017. Accessed: November 19, 2018. URL: https://www.oslobodjenje.ba/vijesti/bih/covic-za-hrt-treci-entitet-je-realna-mogucnost- trebali-smo-ga-traziti-prije-20-godina

Pabriks, Artis & Aiga Stokenberga. 2006. Political Parties and the Party System in Latvia. In Post-communist EU Member States: Parties and Party Systems, ed. Susanne Jungerstam- Mulders. Burlington: Ashgate pp. 51–67.

Pande, Rohini. 2003. “Can mandated political representation increase policy inﬂuence for disadvantaged minorities? Theory and evidence from India.” American Economic Review 93(4):1132–1151.

Pandora’s Box. 2018. The Economist 428(9107):26. URL: http://search.proquest.com/docview/2097948801/abstract/ D1D7D4E1CF3F4077PQ/1

Pasic, Lana. 2013. “How Many are We? Bosnia’s First Post-War Census.”. October 1, 2013, Accessed: October 3, 2018. URL: "https://www.aljazeera.com/indepth/opinion/2013/10/how-many-are-bosnia-ﬁrst-post- war-census-2013101131831546241.html"

Pauker, Iva. 2012. War Through Other Means: Examining the Role of Symbols in Bosnia and Herzegovina. In Peace Psychology in the Balkans: Dealing with a Violent Past while Building Peace, ed. Olivera Simić, Zala Volčič & Catherine R. Philpot. Peace Psychology Book Series Boston: Springer pp. 109–128.

Penikis, Andrejs. 1996. “The Third Awakening Begins: The Birth of the Latvian Popular Front, June 1988 to August 1988.” Journal of Baltic Studies 27(4):261–290.

Perry, Valery. 2013. “The 2013 Census in Bosnia and Herzegovina–a Basic Review.” DPC Policy notes Series 3.

Pilet, Jean-Benoit & Emilie van Haute. 2012. Criteria, Conditions, and Procedures for Establishing a Political Party in the Member States of the European Union. Technical report European Parliament. Directorate-General for Internal Policies. Policy Department: Citizens’ Rights and Constitutional Aﬀairs Brussels: . Accessed: November 5, 2018. URL: http://www.europarl.europa.eu/RegData/etudes/etudes/join/2012/462512/IPOL- AFCO_ET(2012)462512_EN.pdf

Plakans, Andrejs. 1974. “Peasants, Intellectuals, and Nationalism in the Russian Baltic Provinces, 1820-90.” The Journal of Modern History 46(3):445–475.

Polk, Jonathan, Jan Rovny, Ryan Bakker, Erica Edwards, Liesbet Hooghe, Seth Jolly, Jelle Koedam, Filip Kostelka, Gary Marks, Gijs Schumacher et al. 2017. “Explaining the Salience of Anti-elitism and Reducing Political Corruption for Political Parties in Europe with the 2014 Chapel Hill Expert Survey data.” Research & Politics 4(1):1–9.

Popkin, Samuel L & Samuel L Popkin. 1994. The Reasoning Voter: Communication and PER- SUAsion in Presidential Campaigns. Chicago: University of Chicago Press.

267 Posavska Stranka. 2010. “Programska Deklaracija.” http://www.posavskastranka. org/index.php?option=com_content&view=article&id=102:programska-deklaracija- ps&catid=40&Itemid=578&showall=&limitstart=1.

Posner, Daniel N. 2004a. “Measuring Ethnic Fractionalization in Africa.” American Journal of Political Science 48(4):849–863.

Posner, Daniel N. 2004b. “The political salience of Cultural Diﬀerence: Why Chewas and Tum- bukas are Allies in Zambia and Adversaries in Malawi.” American Political Science Review 98(4):529–545.

Prazauskas, Algis. 1991. “Ethnic Conﬂicts in the Context of Democratizing Political Systems.” Theory and Society 20(5):581–602.

Šedo, Jakub. 2010. Party System of Bosnia-Herzegovina. In Party Politics in the Western Balkans, ed. Vera Stojarova & Peter Emerson. New York: Routledge pp. 85–98.

Protsyk, Oleh & Stela Garaz. 2013. “Politicization of Ethnicity in Party Manifestos.” Party Politics 19(2):296–318.

Purs, Aldis. 2012. Baltic Facades : Estonia, Latvia and Lithuania since 1945. London: Reaktion Books.

Radio Liberty. 2017. “U.S. Imposes Sanctions On Republika Srpska’s President Dodik.” http: //www.rferl.org/a/dodik-republika-srpska-united-states-sanctions/28239895.html.

Rae, Douglas W. 1967. The Political Consequences of Electoral Laws. New Haven: Yale University Press.

Ramet, Sabrina P. 2012. “The ICTY–Controversies, Successes, Failures, Lessons.” Southeastern Europe 36(1):1–9.

Ramet, Sabrina Petra. 2002. Balkan Babel: The Disintegration of Yugoslavia from the Death of Tito to the fall of Milosevic. Boulder: Routledge.

Rascoe, Ayesha. 2019. “Twitter Analysis Shows How Trump Tweets Diﬀerently About Nonwhite Lawmakers.” NPR Morning Edition . October 10, 2019. Accessed October 29, 2019. URL: https://www.npr.org/2019/10/10/768646968/as-summer-heated-up-trumps-tweets- about-non-white-democrats-intensiﬁed

Reilly, Ben. 2001. Democracy in Divided Societies: Electoral Engineering for Conﬂict Man- agement. Cambridge, U.K. ; New York: Cambridge University Press.

Rich, Vera. 1987. “Soviet Writers Protest Successfully about Environment Science.” Nature p. 455.

Riker, William H. 1962. The Theory of Political Coalitions. New Haven: Yale University Press.

Riker, William H & Peter C Ordeshook. 1968. “A Theory of the Calculus of Voting.” American Political Science Review 62(1):25–42.

268 Roman, Leslie G. 1993. White is a Color! White Defensiveness, Postmodernism, and Anti- racist Pedagogy. In Race, Identity and Representation in Education, ed. Cameron McCarthy & Warren Crichlow. Routledge New York pp. 71–88.

Rose, Richard & Neil Munro. 2003. Elections and Parties in new European Democracies. Washington, DC: CQ Press.

Rustow, Dankwart A. 1970. “Transitions to Democracy: Toward a Dynamic Model.” Compar- ative politics 2(3):337–363.

Saarts, Tõnis. 2011. “Comparative Party System Analysis in Central and Eastern Europe: the Case of the Baltic States.” Studies of Transition States and Societies; Tallin 3(3):83–104.

Sabanovic, Ala. N.d. “The Polish National Minority in Lithuania: Three Reports Later.” Nouvelle Europe. Forthcoming. January 3, 2012. Accessed: October 29, 2019. URL: http://www.nouvelle-europe.eu/en/polish-national-minority-lithuania-three-reports- later

Sakwa, Richard. 1998. “Left or Right? The CPRF and the Problem of Democratic Consolidation in Russia.” The Journal of Communist Studies and Transition Politics 14(1-2):128–158.

Sartori, Giovanni. 1976. Parties and Party Systems: A Framework for Analysis. Cambridge: Cambridge University Press.

Savezni zavod za statistiku. 1981. Statistički godišnjak Jugoslavije. Technical report Socijalis- tička Federativna Republika Jugoslavija Belgrade: .

Schermerhorn, Richard Alonzo. 1978. Ethnic Plurality in India. Tuscon: Univ of Arizona Press.

Schulze, Jennie L. 2018. Strategic Frames: Europe, Russia, and Minority Inclusion in Estonia and Latvia. Pittsburgh: University of Pittsburgh Press.

Sciolino, Elaine. 1994. “U.S. Says Force Is Still a Choice In Bosnia Crisis.” The New York Times p. A1.

Selimbegović, Vildana. 2001. “Sefer Halilović, haški optuženik.” Bosanskohecegovački Dani . September 28, 2001, Accessed: February 2, 2014. URL: http://bhdani.com/arhiva/225/t22512.shtml

Sheeter, Laura. 2006. “’No Swastikas’ For Nato Mittens.” BBC News . August 20, 2006. Ac- cessed June 18, 2019. URL: http://news.bbc.co.uk/2/hi/europe/5268950.stm

Shiveley, W. Phillips. 1980. The Craft of Political Research. Upper Saddle River, NJ: Prentice Hall.

Shrader, Charles R. 2003. The Muslim-Croat Civil War in Central Bosnia: A Military History, 1992-1994. Vol. 23 Texas A&M University Press.

Shtromas, Aleksandras. 1999. The Baltic States as Soviet Republics. In The Baltic States: the national self-determination of Estonia, Latvia and Lithuania, ed. Graham Smith. 1. published, digital printing ed. Basingstoke, Hampshire: Macmillan pp. 86–117. OCLC: 248347318.

269 Sisk, Timothy D & Andrew Reynolds. 1998. Elections and Conﬂict Management in Africa. Washington: US Institute of Peace Press.

Smith, Anthony D. 1986. The Ethnic Origins of Nations. Oxford: Basil Blackwell Oxford.

Smith, Graham. 1996. “The Ethnic Democracy Thesis and the Citizenship Question in Estonia and Latvia.” Nationalities Papers 24(2):199–216.

Smith-Sivertsen, Hermann. 2004. Latvia. In The Handbook of Political Change in Eastern Europe, ed. Sten Berglund, Joakim Ekman & Frank Aarebrot. 2. ed ed. Cheltenham: Elgar p. Chapter 4. OCLC: 845615929.

Sobolewska, Maria. 2013. “Party Strategies and the Descriptive Representation of Ethnic Mi- norities: The 2010 British General Election.” West European Politics 36(3):615–633.

Socijaldemokratska Partija BiH. 2014. “Program Socijaldemokratske partije Bosne i Herce- govine.” http://www.sdp.ba/upload/documents/dokumenti/PROGRAM_SDP_BiH.pdf.

Special Issue: Albanian Migration and New Transnationalisms. 2003. Journal of Ethnic and Migration Studies 29(6).

Steger, Tamara Shevaun. 2004. Environmentalism and Democracy in Hungary and Latvia PhD thesis Syracuse University.

Stepan, Alfred. 2015. “India, Sri Lanka, and the Majoritarian Danger.” Journal of Democracy 26(1):128–140.

Stepan, Alfred, Juan J Linz & Yogendra Yadav. 2011. Crafting State-Nations: India and Other Multinational Democracies. Baltimore: JHU Press.

Stoll, Heather. 2013. Changing Societies, Changing Party Systems. Cambridge: Cambridge University Press.

Stranka Demokraske Akcije. 2014. “Programska Deklaracija SDA.” http://sda.ba/home/wp- content/uploads/2015/06/PROGRAMSKA-DEKLARACIJA-SDA.pdf.

Surk, Barbara. 2018. “Milorad Dodik Wants to Carve Up Bosnia. Peacefully, if Possible.” The New York Times . February 16, 2018. Accessed: July 14, 2019. URL: https://www.nytimes.com/2018/02/16/world/europe/dodik-republika-srpska-bosnia. html

SzBiH. 2015. “Programska Orijentacija Stranke za Bosnu i Hercegovinu.”. URL: http://zabih.ba/wp-content/uploads/2017/09/PROGRAMSKA-ORIJENTACIJA- SBIH-ﬁnal.pdf

Szöcsik, Edina & Christina Isabel Zuber. 2015. “EPAC–a New Dataset on Ethnonationalism in Party Competition in 22 European Democracies.” Party Politics 21(1):153–160.

Taagepera, Mare. 1981. “Pollution of the Environment and the Baltics.” Journal of Baltic Studies 12(3):260–274.

Taagepera, Rein & Matthew Soberg Shugart. 1989. Seats and Votes: The Eﬀects and Determi- nants of Electoral Systems. New Haven: Yale University Press.

270 Tajfel, Henri. 1978. Social Categorization, Social Identity and Social Comparison. In Diﬀer- entiation Between Social Groups: Studies in the Social Psychology of Intergroup Relations, ed. Henri Tajfel. Number 14 in “European Monographs in Social Psychology” London ; New York: Academic Press.

Tajfel, Henri. 1982. Social Identity and Intergroup Relations. Cambridge: Cambridge Univer- sity Press.

Tavits, Margit. 2007. “Clarity of Responsibility And Corruption.” American Journal of Political Science 51(1):218–229.

Tavits, Margit. 2008. “Party Systems in the Making: The Emergence and Success of New Parties in new Democracies.” British Journal of Political Science 38(1):113–133.

Teney, Celine, Dirk Jacobs, Andrea Rea & Pascal Delwit. 2010. “Ethnic Voting in Brussels: Voting Patterns among Ethnic Minorities in Brussels (Belgium) during the 2006 Local Elec- tions.” Acta Politica 45(3):273–297.

The Death of Yugoslavia. 1995. Film. United Kingdom: Brian Lapping Associates. Percy, Norma [Executive Producer].

Thompson, Mark. 1992. A Paper House : the Ending of Yugoslavia. London: Hutchinson Radius.

Toal, Gerard. 2013. ““Republika Srpska will have a Referendum”: the Rhetorical Politics of Milorad Dodik.” Nationalities Papers 41(1):166–204.

Toe, Rodolfo. 2016. “Bosnia to Publish Census Without Serb Agreement.”. June 29, 2016. Accessed: October 3, 2018. URL: "http://www.balkaninsight.com/en/article/bosnia-to-release-long-awaited-census- results-on-thursday-06-29-2016"

Turner, John C, Michael A Hogg, Penelope J Oakes, Stephen D Reicher & Margaret S Wetherell. 1987. Rediscovering the Social Group: A Self-Categorization Theory. Oxford: Basil Black- well.

Turnham, Steve. 2016. “Donald Trump to Father of Fallen Soldier: ’I’ve Made a Lot of Sacriﬁces’.” ABC News . June 30, 2016. Accessed: October 29, 2019. URL: https://abcnews.go.com/Politics/donald-trump-father-fallen-soldier-ive-made- lot/story?id=41015051

US Census Bureau. 2018. “2017 ACS 1-year Estimates.”. accessed July 29, 2019. URL: https://www.census.gov/programs-surveys/acs/technical-documentation/table-and- geography-changes/2017/1-year.html

Van Cott, Donna Lee. 2005. From Movements to Parties in Latin America : the Evolution of Ethnic Politics. Cambridge: Cambridge University Press. van der Wilt, Harmen. 2007. “Halilović on Appeal: The Intricate Concept of ‘Eﬀective Control’.” The Hague Justice Portal . URL: http://www.haguejusticeportal.net/Docs/Commentaries%20PDF/Van%20der% 20Wilt%20Final.pdf

271 Vanajuur, Otto-Kristjan. 2019. “Estonian Elections: Splits in the Centre Ground.” Europe Elects . February 25, 2019. Accessed: October 29, 2019. URL: https://europeelects.eu/2019/02/25/estonian-elections-splits-in-the-centre-ground/

Velikonja, Mitja. 2003. Religious Separation and Political Intolerance in Bosnia-Herzegovina. College Station: Texas A&M University Press.

Volkens, Andrea, Pola Lehmann, Theres Matthieß, Nicolas Merz, Sven Regel & Bernhard Weßels. 2017. “The Manifesto Data Collection. Manifesto Project (MRG/CMP/MARPOR). Version 2017b.”. URL: https://doi.org/10.25522/manifesto.mpds.2017b

Vrbek, Sanja. 2015. “Reasons for EU Double Standards: Comparative Overview of the Cases of the Erased and the Non-citizens.” Nationalities Papers 43(2):302–318.

Vukelic, Kristijan. 2012. “Designing Special Electoral Mechanisms for Group Representation.” MA Thesis. Central European University.

Weber, Eugen. 1976. Peasants into Frenchmen: the Modernization of Rural France, 1870-1914. Stanford, CA: Stanford University Press.

Wedeen, Lisa. 2002. “Conceptualizing Culture: Possibilities for Political Science.” American Political Science Review 96(4):713–728.

Wedeen, Lisa. 2010. “Reﬂections on Ethnographic Work in Political Science.” Annual Review of Political Science 13:255–272.

White, Harrison C. 1992. Identity and Control: a Structural Theory of Social Action. Princeton, N.J.: Princeton University Press.

Wilson, Stacey-Ann. 2012. Politics of Identity in Small Plural Societies: Guyana, the Fiji Is- lands, and Trinidad and Tobago /. 1st ed. ed. New York: Palgrave Macmillan,.

Wimmer, Andreas. 2008. “The Making and Unmaking of Ethnic Boundaries: A Multilevel Process Theory.” American Journal of Sociology 113(4):970–1022.

Wimmer, Andreas. 2013. Ethnic Boundary Making: Institutions, Power, Networks. New York: Oxford University Press.

Zake, I. 2002. “The People’s Party in Latvia: Neo-Liberalism and the New Politics of Indepen- dence.” Journal of Communist Studies and Transition Politics 18(3):109–131.

Zake, Ieva. 2007. “Inventing Culture and Nation: Intellectuals and Early Latvian Nationalism.” National Identities 9(4):307–329.

Zakharov, Nikolay & Ian Law. 2017. Post-Soviet Racisms. Springer.

Zdeb, Aleksandra. 2016. “The Need to Have Something ‘of their own’: Croat Parallel Institu- tions in Bosnia and Herzegovina.” Swiss Political Science Review 22(4):545–564.

Zulianello, Mattia. 2014. “Analyzing Party Competition through the Comparative Manifesto Data.” Quality & Quantity 48(3):1723–1737.

272